POINT CLOUD ATTRIBUTE ENCODING AND DECODING METHOD AND DEVICE

TECHNICAL FIELD

The present disclosure belongs to the field of point cloud coding, and particularly relates to a point cloud attribute decoding method and device based on a new processing order.

BACKGROUND ART

Supporting the application of virtual reality (VR) AR, Augmented Reality (AR) is the next important challenge of media technology in the 21st century. These applications, together with 3D data acquisition, 3D rendering and computer vision, make it possible to realize the emerging immersive media experience in the near future. Point cloud is an ideal representation of three-dimensional scenes and objects. It consists of a group of points that are irregularly distributed and unconnected in three-dimensional space. Each point that constitutes a point cloud has geometric information and optional attribute information. Among them, geometric information is used to express the structure of three-dimensional objects, which is composed of three-dimensional position coordinates X, Y, and Z; Attribute information is used to express the surface attributes of three-dimensional scenes or objects. Common attribute information includes color, reflectivity, normal, and transparency. The data scale of point cloud is extremely huge. For example, a point cloud with high quality to express three-dimensional objects has hundreds of thousands to millions of points. In order to facilitate the storage and transmission of point cloud data, it is necessary to compress point cloud data.

Most point cloud compression algorithms are based on voxelized point cloud, which has the advantages of convenient processing and controllable accuracy. A voxelized point cloud is a point cloud whose pointing position coordinates are quantized into a cube grid, and the coordinates of the midpoint of each cube grid represent the coordinates of the midpoint of the cube grid, so only the index of the cube grid is needed to encode the position information of the point cloud, and the position accuracy can be controlled by the size of the cube grid.

At present, most point cloud attribute coding is based on Morton order to find several nearest neighbors, and then the attributes of these nearest neighbors are weighted to predict the attribute value of the current point. Among them, Morton order is the point order obtained by sorting the points in the point cloud according to their Morton codes, and the Morton codes of each point can be obtained from the position coordinates of the points. Specifically, it is obtained by alternately placing the coordinates of each dimension of the points according to each bit of the binary bit. In [1], the nearest neighbor points are to be found based on Morton order. If the number of LOD (level of detail) is 1, three points closest to the current point will be found from several coded points before the current point under Morton order, and then the weighted average of the attribute values of these points will be used to predict the attribute values of the current point. If the number of LOD (level of detail) is greater than 1, Then the first point whose Morton code is greater than the Morton code of the current point will be found from the points in the coded level of detail arranged in Morton order, and then three points closest to the current point will be found around this found point, and the weighted average of the attribute values of these points will be used to predict the attribute values of the current point. In [2], attribute prediction is also based on Morton code. For sparse point clouds, the weighted average of the attribute values of the first three points of the current point in Morton order is directly used to predict the attribute values of the current point. For dense point clouds, the points that are coplanar and collinear with the current point are first found from the first several points of the current point in Morton order. If they are found, the weighted average of the attribute values of these coplanar and collinear points is used to predict the attribute values of the current point, otherwise, the previous points of the current point in Morton order are directly used.

Theoretically, the closer the three-dimensional distance between the predicted point and the current point is, the stronger the correlation between the predicted point's attributes and the current point's attributes is, the more accurate the attribute prediction is, and the lower the required code rate under the condition of constant distortion is. Since the three-dimensional spatial distance between adjacent or similar points under the Morton order will periodically jump with different amplitudes, the three-dimensional spatial distance between adjacent points under the Morton order will periodically change from near to far. Therefore, the Morton code has poor spatial neighbor characteristics, so the nearest point found by the nearest neighbor search with the Morton order will not be ideal.

Hilbert curve is a continuous fractal space filling curve, also known as Hilbert space filling curve. It was first discovered by German mathematician David Hilbert in 1891, and it can map high-dimensional space to one dimension. Hilbert curve has good spatial neighborhood characteristics. Hilbert's spatial neighborhood characteristics mean that data points that are similar in one-dimensional space are also similar in high-dimensional space. Because of its spatial neighborhood, Hilbert curve and its discrete approximation are widely used. For example, it is widely used in database indexing, memory management, image processing, and dynamic load balancing. This disclosure uses the discrete approximation of Hilbert curve in three-dimensional space.

SUMMARY

For the shortcoming that the three-dimensional space distance before adjacent points in Morton order will periodically jump, the present disclosure proposes a point cloud attribute decoding method and device based on a new processing order to further improve the coding efficiency of point cloud attributes. Hilbert order is used as the coding order of point cloud attribute. The nearest neighbors in the three-dimensional space of the current point are found from several points in the preceding of the current point in Hilbert order, and these nearest neighbors are used as the prediction points of the current point. Because the similar points in Hilbert order are also similar in three-dimensional space, in particular, in the worst case, the distance between two points in three-dimensional space with distance n on Hilbert order is less than or equal to and the distance between two points on Morton order with distance n in three-dimensional space may be greater than the maximum value of point cloud dimension, and when k is small, the average distance between the current point and the current point in Hilbert order is smaller than that of the current point in Morton order. Therefore, Hilbert order can be used to find the neighbor points that are more similar to the current point in three-dimensional space, so that the attribute values of the predicted points are more relevant to the attribute values of the current point, and then the coding efficiency of point cloud attributes can be improved.

The first purpose of the disclosure is to provide a point cloud attribute decoding method based on a new processing order, which includes the following contents:

- The weighted average of the reconstructed attribute values of the preceding decoding points under Hilbert order is used as the attribute prediction value of the current point, wherein Hilbert order is the point order obtained by sorting the points in the point cloud from small to large or from large to small according to their corresponding Hilbert code values;
- Decoding the attribute residual value of the current point from the bitstream;
- According to the attribute prediction value of the current point and the decoded attribute residual value, the reconstructed attribute value of the current point is obtained.

Preferably, the point cloud attribute decoding method based on the new processing order is characterized in that the weighted average of the reconstructed attribute values of the preceding decoding points under Hilbert order comprises:

- N decoding points in the first order under Hilbert order are used as the attribute prediction points of the current point, where n is a natural number;
- The weighted average of the reconstructed attribute values of these n attribute prediction points is used as the attribute prediction value of the current point.

Preferably, the point cloud attribute decoding method based on the new processing order is characterized in the weighted average of the reconstructed attribute values of the preceding decoding points under Hilbert order comprises:

- Use n points closest to the current point from L points in the preceding of Hilbert order as the attribute prediction points of the current point, where L is a natural number greater than 0 and n is a natural number;
- The weighted average of the reconstructed attribute values of these n attribute prediction points is used as the attribute prediction value of the current point.

- Using coplanar points of the current point and collinear points of the current point among L numbered preceding points in the Hilbert order as prediction points of attribute of the current point, wherein each of the coplanar points is the point located in its cube that has a common surface with the cube where the current point is located, each of the collinear points is the point located in its cube that has a common edge with the cube where the current point located, and L is a natural number greater than 0;
- Weight coplanar points and collinear points ω_pland ω_po, in which ω_pland ω_poare real numbers greater than or equal to 0, ω_plis greater than ω_po;
- The weighted average of reconstructed attribute values from coplanar points and collinear points is used as the attribute prediction value of the current point.

Preferably, the point cloud attribute decoding method based on the new processing order can also have another implementation form, which is characterized by comprising:

- Constructing L-level level of detail (LOD_i)_{i=0,1 . . . L−1}, among which, LOD_iis a subset of the reconstructed point cloud, LOD_i⊂LOD_j(i<j), LOD_L−1is the whole reconstructed point cloud, where L is a natural number greater than 0;
- Obtain n attribute prediction points of the current point, and make (R_i=LOD_i−LOD_i−1)_{i=1 . . . L−1}, R₀=LOD₀, for a point P of Hilbert code H in R_i, Select n points closest to point P from LOD_i−1Hilbert code in the range of [H₁, H₂](H₁≤H≤H₂). The nearest n points as the attribute prediction points of the current point, where n is a natural number;
- The weighted average of the reconstructed attribute values using the n attribute prediction points is taken as the attribute prediction value of the current point;
- Decoding the attribute residual value of the current point from the bitstream;
- According to the attribute prediction value of the current point and the decoded attribute residual value, the reconstructed attribute value of the current point is obtained.

The expression method in the present disclosure is agreed as follows:

- 1. The 3D geometric coordinates of points in the point cloud are marked as (X₁, X₂, X₃)_N, where X₁, X₂, X₃are respectively expressed by binary expansion as follows:

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=

(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

The left side of binary expansion represents high-bit and the right side represents low-bit, where N is an integer greater than 0, indicating the coordinate bit width.

- 2. Hilbert code of point HCode_Nrepresented in binary as C_1,m₁C_2,m₂. . . C_i,m_i. . . C_p-1,m_p−1C_p,m_p, in which the Hilbert code subcode C_i,m_iis binary number of m_ibits.

Preferably, the point cloud attribute decoding method based on the new processing order includes the following steps:

The three-dimensional geometric coordinates of reconstructed points in the point cloud are processed sequentially from high-bit to low-bit. (X₁, X₂, X₃)_NThe binary bits of, where the firstiSecondary processing: according to the three-dimensional geometric coordinates of the reconstructed point (3*m_i) Binary bits, generate the corresponding Hilbert code subcode C_i,m_i:

Process sequentially binary bits of the three-dimensional geometric coordinates (X₁, X₂, X₃)_Nof reconstructed points in the point cloud from high-bit to low-bit, where the i^thtime processing: generate the corresponding subcode C_i,m_iof Hilbert code according to (3*m_i) binary bits of the three-dimensional geometric coordinates of the reconstructed point:

C
_i,m
_i=ConvertCode^mⁱ(x₁^m¹^+m²^{. . . +m}ⁱ⁻¹⁺¹x₁^m¹^+m²^{. . . +m}ⁱ⁻¹⁺². . . x₁^m¹^+m²^{. . . +m}ⁱ,

x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺¹
x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺²
. . . x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ,

x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺¹
x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺²
. . . x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ,state_i),

state_i+1=ConvertState^mⁱ(x₁^m¹^+m²^{. . . +m}ⁱ⁻¹⁺¹x₁^m¹^+m²^{. . . +m}ⁱ⁻¹⁺². . . x₁^m¹^+m²^{. . . +m}ⁱ,

x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺¹
x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺²
. . . x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ,

x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺¹
x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺²
. . . x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ,state_i),

Among them, ConvertCode^mⁱfunction generates the corresponding Hilbert code subcode according to the status state of the binary bit and the binary bit in this processing. There are different ConvertCode^mⁱwhen m_iis different. The mechanism of generating Hilbert code subcodes by ConvertCode^mⁱis based on the shape of the Hilbert sub-curve determined by state and the serial number on the Hilbert sub-curve determined by the binary bits in this processing ConvertState^mⁱfunction generates the state state_i+1of the binary bit in the next processing according to the status state of the binary bit and the binary bit in this processing. ConvertCode^mⁱis different when m_iis different. The mechanism that ConvertState^mⁱgenerates the state state_i+1of binary bit in the next processing according to the shape of the Hilbert sub-curve determined by state and the serial number on the Hilbert sub-curve determined by the processed binary bit. The state corresponding to the binary bit corresponds to the shape of the Hilbert sub-curve corresponding to the binary bit, and a total of p treatments were performed, and the sum of m_iof each treatment is n, that is

Σ_i=1^pm_i=N;

According to the Hilbert order from small to large or from large to small, the reconstruction attribute information of point cloud is decoded from the bitstream in order.

A second object of the present disclosure is to provide a point cloud attribute decoding device based on a new processing order, which includes:

- An attribute prediction value acquisition module, wherein the input of the module is the reconstructed point cloud, and the output is the attribute prediction value of the current point, and the module uses the weighted average of the reconstructed attribute values of the preceding decoding points in Hilbert order as the attribute prediction value of the current point, wherein the Hilbert order is the point order obtained by sorting the points in the point cloud from small to large or from small to small according to their corresponding Hilbert code values;
- A module of decoding attribute residual: the input of this module is the bitstream and the output is the attribute residual, and this module decodes the attribute residual value of the current point from the bitstream;
- A module of acquiring reconstructed attribute value: the input of this module is the attribute residual and the attribute prediction value, and the output is the reconstructed attribute value of the current point, and the module obtains the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value.

Preferably, the point cloud attribute decoding device based on the new processing order is characterized in that the weighted average of the reconstructed attribute values of the preceding decoding points under Hilbert order comprises:

- A module of acquiring attribute prediction point: the input of the module is the reconstructed point cloud, and the output is the attribute prediction point of the current point, and the module uses N decoded points under Hilbert order as the attribute prediction points of the current point, where n is a natural number;
- A module of calculating attribute prediction value: the input of this module is the attribute prediction point, and the output is the attribute prediction value of the current point, and the module uses the weighted average of the reconstructed attribute values of these n attribute prediction points as the attribute prediction value of the current point.

Preferably, the point cloud attribute decoding device based on the new processing order is characterized in the weighted average of the reconstructed attribute values of the preceding decoding points under Hilbert order comprises:

- A module of acquiring attribute prediction point: the input of the module is the reconstructed point cloud, and the output is the attribute prediction point of the current point, and the module uses n points closest to the current point in the top L points of Hilbert order as the attribute prediction points of the current point, where L is a natural number greater than 0 and N is a natural number;
- A module of calculating attribute prediction value: the input of this module is the attribute prediction point, and the output is the attribute prediction value of the current point, and the module uses the weighted average of the reconstructed attribute values of these n attribute prediction points as the attribute prediction value of the current point.

- A module of acquiring attribute prediction point: the input of the module is the reconstructed point cloud, and the output is the attribute prediction point of the current point, and the module uses the coplanar points and collinear points of the current point among the preceding L points in the Hilbert order as the prediction points of attribute of the current point, wherein each of the coplanar points is the point located in its cube that has a common surface with the cube where the current point is located, and each of the collinear points is the point located in its cube that has a common edge with the cube where the current point is located, and L is a natural number greater than 0;
- A module of weighting attribute prediction point: the input of this module is prediction point of attribute, and the output is prediction point of attribute with weight, and this module gives coplanar points and collinear points weights ω_pland ω_po, in which ω_pland ω_poare real number greater than or equal to 0, ω_plis bigger than ω_po;
- A module of calculating attribute prediction value: the input of this module is the weighted attribute prediction point, and the output is the attribute prediction value of the current point. This module uses the weighted average of the reconstructed attribute values of coplanar points and collinear points as the attribute prediction value of the current point.

Preferably, the device of decoding point cloud attribute based on the new processing order can also have another implementation form, whose characteristics include:

- A module of constructing LOD: the input of this module is the reconstructed point cloud geometry, and the output of this module is the constructed LOD. This module builds (LOD_i)_{i=0,1 . . . L−1}as L-level detail level, among which, LOD_iis a subset of the reconstructed point cloud, LOD_i⊂LOD_j(i<j), LOD_L−1is the whole reconstructed point cloud, where L is a natural number greater than 0;
- A module of acquiring attribute prediction point: the input of the module is the constructed LOD, and the output is the attribute prediction point of the current point, and the module acquires n attribute prediction points of the current point, make (R_i=LOD_i−LOD_i−1)_{i=1 . . . L−1}, R₀=LOD₀, for a point P of Hilbert code H in R_i, Select n points closest to point P from LOD_i−1Hilbert code in the range of [H₁, H₂](H₁≤H≤H₂). The nearest n points as the attribute prediction points of the current point, where n is a natural number;

A module of calculating attribute prediction value: the input of this module is the attribute prediction point, and the output is the attribute prediction value of the current point, and the module uses the weighted average of the reconstructed attribute values of these n attribute prediction points as the attribute prediction value of the current point;

A module of decoding attribute residual: the input of this module is the bitstream and the output is the attribute residual, and this module decodes the attribute residual value of the current point from the bitstream;

A module of acquiring reconstructed attribute value: the input of this module is the attribute residual and the attribute prediction value, and the output is the reconstructed attribute value of the current point, and the module obtains the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value.

Preferably, the method of decoding point cloud attribute based on the new processing order includes the following steps:

A module of generating Hilbert code: the input of this module is the reconstructed geometric coordinates of the point cloud, and the output is the Hilbert code corresponding to the geometric coordinates of the point cloud. This module sequentially processes binary bits of the three-dimensional geometric coordinates (X₁, X₂, X₃)_Nof reconstructed points in the point cloud from high-bit to low-bit, where the ith time processing: generate the corresponding subcode C_i,m_iof Hilbert code according to (3*m_i) binary bits of the three-dimensional geometric coordinates of the reconstructed point:

C
_i,m
_i=ConvertCode^mⁱ(x₁^m¹^+m²^{. . . +m}ⁱ⁻¹⁺¹x₁^m¹^+m²^{. . . +m}ⁱ⁻¹⁺². . . x₁^m¹^+m²^{. . . +m}ⁱ,

x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺¹
x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺²
. . . x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ,

x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺¹
x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺²
. . . x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ,state_i),

state_i+1=ConvertState^mⁱ(x₁^m¹^+m²^{. . . +m}ⁱ⁻¹⁺¹x₁^m¹^+m²^{. . . +m}ⁱ⁻¹⁺². . . x₁^m¹^+m²^{. . . +m}ⁱ,

x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺¹
x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺²
. . . x
₂
^m
¹
^+m
²
^{. . . +m}
ⁱ,

x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺¹
x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ⁻¹
⁺²
. . . x
₃
^m
¹
^+m
²
^{. . . +m}
ⁱ,state_i),

ConvertCode^mⁱfunction generates the corresponding subcode of Hilbert code according to the state state_iof the binary bit and the binary bit in this processing. ConvertCode^mⁱis differernt when m_iis different. The mechanism that ConvertCode^mⁱgenerates subcode of Hilbert code is: generate the corresponding subcode of Hilbert code according to the shape of the Hilbert sub-curve determined by state state_iand the serial number on the Hilbert sub-curve determined by the binary bits in this processing. ConvertState^mⁱfunction generates the state state_i+1of the binary bit in the next processing according to the state state_iof the binary bit and the binary bit in this processing. ConvertCode^mⁱis diffferent when m_iis different. The mechanism that ConvertState^mⁱgenerates the state state_i+1of binary bit in the next processing is: to determine the state state_i+1of binary bit in the next processing according to the shape of the Hilbert sub-curve determined by state and the serial number on the Hilbert sub-curve determined by the binary bit in this processing. The state corresponding to the binary bit corresponds to the shape of the Hilbert sub-curve corresponding to the binary bit one by one, and a total of p processings are performed, and the sum of each processing m_iis n, that is

Σ_i=1^pm_i=N;

A module of decoding attribute: the input is the attribute bitstream, and the output is the reconstructed point cloud attribute in this module. The module sequentially decodes the bitstream to get the reconstructed point cloud attribute information according to the order of Hilbert from small to large or from large to small.

Because the used Hilbert order has the best spatial neighbor characteristics, neighbor points closer to the current point in three dimensions can be found with greater probability as the prediction points. Thus the attribute values of the prediction points are more relevant to the attribute values of the current point. Then we achieve the goal to improve the compression efficiency of point cloud attributes on the premise of maintaining the complexity.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to explain the technical scheme in the embodiment of the application or the prior art more clearly, the following will briefly introduce the drawings needed in the description of the embodiment or the prior art Obviously, the drawings in the following description form a part of this application, and the illustrative embodiments of the disclosure and their descriptions are used to explain the disclosure, and do not constitute an improper limitation of the disclosure.

For ordinary technicians in this field, other drawings can be obtained according to these drawings without creative work.

FIG. 1 is a schematic diagram of a three-dimensional Hilbert curve according to the present disclosure.

FIG. 2 is an algorithm flow chart of an embodiment of the present disclosure.

FIG. 3 is a schematic diagram of the device according to the patent embodiment of the disclosure.

FIG. 4 shows 12 Hilbert states related to the present disclosure.

FIG. 5 is a schematic diagram of the device according to the patent embodiment of the disclosure.

DETAILED DESCRIPTION OF THE DISCLOSURE

In order to further understand the present disclosure, the preferred embodiments of the present disclosure are described below with reference to examples, but it should be understood that these descriptions are only for further explaining the features and advantages of the present disclosure, and are not a limitation of the claims of the present disclosure.

For the convenience of expression, the following definitions are made:

For n binary numbers X₁, X₂, . . . , X_n, X₁X₂. . . X_nmeans to put this n binary numbers are bitwise spliced together, and the number on the left is high and the number on the right is low.

If each dimension of a D-dimensional hypercube is divided equally by 2ⁿ, then the hypercube is called a D-dimensional n-order hypercube, expressed by HC_n^D, let (X₁, X₂. . . X_d. . . X_D) denote the coordinates of the smallest sub-cube in D-dimensional n-order hyper-cube, where, 0≤X_d<2ⁿ, 1≤d≤D. A curve passing through the smallest sub-cube in HC_n^Dis called a D-dimensional n-order Hilbert curve denoted by H_n^D. The scan order index of each minimal sub-cube over H_n^Din a D-dimensional n-order hypercube HC_n^Dis called the Hilbert code of the minimal sub-cube, which is represented by HCode_n^D, where HCode_n^Dstarts from 0.

The order of scanning the smallest cube of the D-dimensional 1-order Hilbert curve in the D-dimensional 1-order hyper-cube H₁^Dis the order of gray codes. For example, for three-dimensional, the gray codes of the sub-cubes with the scanning order of 0,1, 2, 3, 4, 5, 6 and 7 are 000, 001, 011, 010, 110, 111, 101 and 100, which are exactly the coordinates of the sub-cubes. Therefore, the Hilbert code of the D-dimensional 1-order curve can be generated according to the gray code.

Symbols (x¹x². . . x^K)_bare used to express K bits binary number, where 0≤x^k<b, 1≤k≤K.

Gray code coding function is

G((x¹x². . . x^K)₂)=(g¹g². . . g^K)₂,

(g¹g². . . g^K)²is the Gray code of (x¹x². . . x^K)₂, the calculation formula is as follows:

(g¹g². . . g^K)²=(x¹x². . . x^K)₂∧((x¹x². . . x^K)²>>1)

Where ∧ denotes exclusive or operators, >> denotes the right shift operator.

The decoding function of Gray code is

DG((g¹g². . . g^K)₂)=(x¹x². . . x^K)₂

Where x^k=DG^k((g¹g². . . g^K)₂)

The calculation formula is as follows

$x^{1} = {DG}_{1} ({(ℊ^{1} ℊ^{2} {…ℊ}^{K})}_{2}) = ℊ^{1},$

$x^{k} = {DG}_{k} ({(ℊ^{1} ℊ^{2} {…ℊ}^{K})}_{2}) = ((\sum_{i = 1}^{k - 1} ℊ^{i}) \mod 2) \land ℊ^{k},$

$2 \leq k \leq K$

Hilbert curve is self-similar, and the generation of the curve is a recursive process. The Hilbert curve of order n+1 needs the Hilbert curve of order N, specifically, the position corresponding to the 2^Dsub-cubes of the D-dimensional 1-order hyper-cube is regarded as the 2^Dquadrant of the D-dimensional space, then for the n+1-order Hilbert curve, it is necessary to copy the n-order Hilbert curve directly or put it in a quadrant of the n+1-order Hilbert curve after mirror flipping, and then connect the starting point and end point of the n-order Hilbert curve of the adjacent quadrant or the n-order Hilbert curve after mirror flipping, so as to form the n+1-order curve. There are two main flips used: mirror flip 1 and mirror flip 2. Both flips are mirror flips along a hyperplane, where

Mirror flip 1 is along the hyperplane X_i−X_j=0 to flip. Exchange the coordinates between X_iand X_jneeded. X_i↔X_jrepresents the exchange of coordinates.

Mirror flip 2 is along the hyperplane

$X_{i} = \frac{2^{n} - 1}{2}$

to flip, the coordinate X_ineeds to be exchanged to be 2ⁿ−1−X_i.

These two flips are order-preserving, that is, the relative position of each point on the curve remains unchanged before and after flipping.

Hilbert sub-curve is directional, and the direction can be uniquely determined by the coordinates of the starting point and the ending point. There is only one different bit between the coordinates of the starting point and the ending point. For example, the starting point coordinate of the three-dimensional first-order Hilbert curve is DG (0)=000, and the ending point coordinate is DG (7)=100. B^qand E^qare used to represent respectively the starting coordinates and ending coordinates of the sub-curves of the D-dimensional 1-order in the quadrant q (as mentioned above, the coordinates of the quadrant q can be obtained by the gray code G(q)). Then for three dimensions, there are B⁰=000, E⁰=010, B¹=000, E¹=100, B²=000, E²=100, B³=110, E³=111, B⁴=110, E⁴=111, B⁵=101, E⁵=001, B⁶=101, E⁶=001, B⁷=001, E⁷=011.

It can be determined by the coordinates of the starting point and the ending point to turn each sub-curve into a line mirror image flip similar to the one-order curve.

It is three-dimensional space that counts in the following. That is: D=3, giving the three-dimensional coordinates (X₁, X₂, X₃) of a point, which is expanded in binary to

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=

(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

The mirror flip 1 of quadrant q

T
₁
^q=(DG(0)∧DG(7))∧(B^q∧E^q),0≤q≤7

The obtained 3 digits T₁^qabove represents the flip rule of mirror flip 1. The specific flip rule is as follows: there is only one different bit between the coordinates of starting point and the coordinates of ending point, so G(0)∧DG(7) and B^q∧E^qonly one digit 1. Therefore, there are only two 1 or all 0s in each binary bit of T₁^q, if the ith bit and jth bit of T₁^qis 1, then exchange coordinates X_iand X_j, 1≤i, j≤3.

The mirror flip 2 of quadrant q

T
₂
^q
=DG(0)∧B^q,0≤q≤7

The obtained 3 digits T₂^qrepresents the flip rule of mirror flip 2. The specific flip rule is as follows: if the ith bit of T₂^qis 1, setting X_i=2ⁿ−1−X_i, where n is the order of the curve to be flipped.

Hilbert coding is the process of calculating the Hilbert code HCode_N=(C₁C₂. . . C_n. . . C_N−1C_N) according to the given coordinates (X₁, X₂, X₃), where C_nis the 3-bit binary subcode of each layer, and n is taken from 1 to N to obtain C_nby the following calculation.

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the ith bit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the ith bit and jth bit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4.

Embodiment 1

A method of decoding point cloud attribute based on a new processing order in this embodiment, as shown in FIG. 2, includes the following contents.

For the reconstructed point cloud geometry, the associated Hilbert code is generated according to its three-dimensional spatial coordinates. As mentioned above, for a point given three-dimensional coordinate (X₁, X₂, X₃), its coordinates are expanded in terms of binary as follows:

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

The Hilbert code of the point (X₁, X₂, X₃) can be obtained by the following process, HCode_N=(C₁C₂. . . C_n. . . C_N−1C_N), where C_nis a 3-bit binary subcode, C_nis obtained by sequentially taking n from 1 to N through the following calculation

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the ith bit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the ith bit and jth bit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4;

The attribute values of points in the point cloud are processed in sequence according to Hilbert order, and the attribute values of the current points are processed by the following steps: Use the weighted average of reconstructed attribute values of preceding decoding points in Hilbert order is as the attribute prediction value of the current point, where Hilbert order is the point order obtained by sorting points in the point cloud from small to large or from large to small according to their corresponding Hilbert code values: use the first point P of the current point in Hilbert order, let its attribute value=Â. That is, Â is the attribute prediction value of the current point;

- 1) Decode the attribute residual value R of the current point from the bitstream;

Obtain the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 2

A method of decoding point cloud attribute based on a new processing order in this embodiment includes: For the reconstructed point cloud geometry, the associated Hilbert code is generated according to its three-dimensional spatial coordinates. As mentioned above, for a point given three-dimensional coordinate (X₁, X₂, X₃), its coordinates are expanded in terms of binary as follows:

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the i^thbit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the i^thbit and j^thbit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4;

The attribute values of points in the point cloud are processed in sequence according to Hilbert order, and the attribute values of the current points are processed by the following steps:

- 1) N proceeding decoding points in Hilbert order are used as the attribute prediction points of the current point, where n is a natural number. n is taken to be 3 in this embodiment: use the first three points of the current 128 points in Hilbert order, P_i=(Z_i, Y_i, Z_i), i=1, 2, 3. Set their attribute values (A_i)_i=1,2,3;
- 2) The weighted average of the reconstructed attribute values from these three attribute prediction points is used as the attribute prediction value of the current point: the weights (ω_i)_i=1,2,3of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

Where d_iis the distance of point P_ifrom current point (X, Y, Z), the distance is at least one of the following distances:

- Euclidean distance: d_i=−1a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value of the current point Â,

$Â = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) Decoding the attribute residual value of the current point from the bitstream R;
- 4) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 3

A point cloud attribute decoding method based on a new processing order in this embodiment includes the following contents.

For the reconstructed point cloud geometry, the corresponding Hilbert code is generated according to its three-dimensional spatial coordinates. As mentioned above, for a given three-dimensional coordinate (X₁, X₂, X₃), the coordinates of the point are expanded in binary as follows:

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

Hilbert code HCode_N=(C₁C₂. . . C_n. . . C_N−1C_N) of the point (X₁, X₂, X₃) can be obtained by the following process, in which C_nis a 3-bit binary subcode, C_nis obtained by taking n from 1 to N in order for the following calculation:

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the i^thbit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the i^thbit and j^thbit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4;

The attribute values of points in the point cloud are processed in sequence according to Hilbert order, and the attribute values of the current points are processed by the following steps:

- 1) N decoding points in the preceding of Hilbert order are selected as the attribute prediction points of the current point, where n is a natural number, and in this embodiment, n is taken as 3: the first three points of the current 128 points under Hilbert order P_i=(X₁, Y_i, Z_i), i=1, 2, 3 are selected. Set their property values as follows (A_i)_i=1,2,3;
- 2) The weighted average of the reconstructed attribute values from these three attribute prediction points is used as the attribute prediction value of the current point: the weights (ω_i)_i=1,2,3of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

- Where, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:
- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value of the current point Â,

$Â = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) Decoding the attribute residual value of the current point from the bitstream R;
- 4) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 4

A point cloud attribute decoding method based on a new processing order in this embodiment includes the following contents.

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the i^thbit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the i^thbit and j^thbit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4;

The attribute values of points in the point cloud are processed in sequence according to Hilbert order, and the attribute values of the current points are processed by the following steps:

- 1) Using coplanar points of the current point and collinear points of the current point among L numbered preceding points in the Hilbert order as prediction points of attribute of the current point, wherein each of the coplanar points is the point located in its cube that has a common surface with the cube where the current point is located, each of the collinear points is the point located in its cube that has a common edge with the cube where the current point located, and L is a natural number greater than 0. In this embodiment, L is taken as 128: search the first 128 points of the current point in Hilbert order for coplanar and collinear points with the current point k, if k is not 0, let these neighbors be (P_i)_{i=1,2 . . . k}, the attribute value of each neighbor point is (A_i)_{i=1,2 . . . k};
- 2) Weight coplanar points and collinear points ω_pland ω_porespectively, in which ω_pland ω_poare real number greater than or equal to 0, ω_plis greater than ω_po, In this embodiment, let ω_pl=2, ω_po=1, the weight of each neighbor point is (w_i)_{i=1,2 . . . k}, in which the neighbor points with a distance of 1 are given a weight of 2, and the neighbor points with a distance of √{square root over (2)} are given a weight of 2;
- 3) The weighted average of the reconstructed attribute value of coplanar points and collinear points is used as the attribute prediction value of the current point: the weighted average of the attribute values of these neighboring points (P_i)_{i=1,2 . . . k}are calculated by using the following formula, and take the weighted average value as the attribute prediction value of the current point Â,

$Â = \frac{\sum_{i = 1}^{k} ω_{i} A_{i}}{\sum_{i = 1}^{k} ω_{i}}$

- 4) Decoding the attribute residual value of the current point from the bitstream R;
- 5) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 5

A point cloud attribute decoding method based on a new processing order in this embodiment includes the following contents.

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the i^thbit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the i^thbit and j^thbit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4;

- Constructing L-level of detail (LOD_i)_{i=0,1 . . . 1−1}, in which, LOD_iis a subset of the reconstructed point cloud, LOD_i⊂LOD_j(i<j), LOD_L−1is the whole reconstructed point cloud, L is a natural number greater than 0, and L is taken as 10 in this embodiment;
- Obtain n attribute prediction points of the current point, and make (R_i=LOD_i−LOD_i−1)_{i=1 . . . L−1}, R₀=LOD₀, for a point P of Hilbert code H in R_i, n points which are closest to the point P are selected from the points of Hilbert code in the range of [H₁, H₂](H₁≤H≤H₂) in LOD_i−1, and these n points are used as the attribute prediction points of the current point, where N is a natural number, in this embodiment, n is taken as 3. Make these three attribute prediction points as P_i=(X₁, Y_i, Z_i), i=1, 2, 3, set their property values as (A_i)_i=1,2,3;
- The weighted average of the reconstructed attribute values from these three attribute prediction points is used as the attribute prediction value of the current point: the weights of the three points (ω_i)_i=1,2,3are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value of the current point Â,

$Â = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- Decoding the attribute residual value of the current point from the bitstream R;
- Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 6

As shown in FIG. 3, a point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules,

Hilbert code generation module: the input of this module is the reconstructed point cloud geometry, and the output is the point cloud geometry containing Hilbert code. For the reconstructed point cloud geometry, this module generates the corresponding Hilbert code according to its three-dimensional spatial coordinates. As mentioned above, for a given three-dimensional coordinate (X₁, X₂, X₃), the coordinates of the point are expanded in binary as follows:

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the i^thbit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the i^thbit and j^thbit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4;

Process the attribute values of points in the point cloud in sequence according to Hilbert order, and execute the following modules to process the attribute values of the current points:

- 1) An attribute prediction value acquisition module: the input of the module is the reconstructed point cloud, and the output is the attribute prediction value of the current point, and the module uses the weighted average of the reconstructed attribute values of the preceding decoding points in Hilbert order as the attribute prediction value of the current point, wherein the Hilbert order is the point order obtained by sorting the points in the point cloud from small to large or from large to small according to their corresponding Hilbert code values: select the first point P of the current point in Hilbert order, set its property values as Â, the property value Â, is the attribute prediction value of the current point;
- 2) Attribute residual decoding module: the input of this module is the bitstream and the output is the attribute residual, and this module decodes the attribute residual value of the current point from the bitstream R;
- 3) Reconstructed attribute value acquisition module: the input of the module is attribute residual and attribute prediction value, and the output is the reconstructed attribute value of the current point, and the module obtains the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value: the reconstructed attribute value of the current point is A=Â+R.

Embodiment 7

A point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules.

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the i^thbit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the i^thbit and j^thbit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4;

Process the attribute values of points in the point cloud in sequence according to Hilbert order, and execute the following modules to process the attribute values of the current points:

- 1) Attribute prediction point acquisition module: the input of this module is the reconstructed point cloud, and the output is the attribute prediction point of the current point. This module selects n decoded points in Hilbert order as the attribute prediction points of the current point, where n is a natural number. In this embodiment, n is taken as 3: the first three points P_i=(X₁, =1, 2, 3 in previous 128 points of the current point in Hilbert order are selected, set their property values as follows (A_i)_i=1,2,3;
- 2) Attribute prediction value calculation module: the input of this module is the attribute prediction point, and the output is the attribute prediction value of the current point. This module uses the weighted average of the reconstructed attribute values of these three attribute prediction points as the attribute prediction value of the current point: the weights (ω_i)_i=1,2,3of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value of the current point Â,

$Â = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) Attribute residual decoding module: the input of this module is the bitstream and the output is the attribute residual, and this module decodes the attribute residual value of the current point from the bitstream R;
- Reconstructed attribute value acquisition module: the input of the module is attribute residual and attribute prediction value, and the output is the reconstructed attribute value of the current point, and the module obtains the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value: the reconstructed attribute value of the current point is A=Â+R.

Embodiment 8

A point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules.

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the i^thbit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the i^thbit and j^thbit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4;

- Process the attribute values of points in the point cloud in sequence according to Hilbert order, and execute the following modules to process the attribute values of the current points:
- 1) Attribute prediction point acquisition module: the input of this module is the reconstructed point cloud, and the output is the attribute prediction point of the current point. This module selects n decoded points in Hilbert order as the attribute prediction points of the current point, where n is a natural number. In this embodiment, n is taken as 3: the first three points P_i=(X_i, Y_i, Z_i), i=1, 2, 3 in previous 128 points of the current point in Hilbert order are selected, set their property values as follows (A_i)_i=1,2,3;
- 2) Attribute prediction value calculation module: the input of this module is the attribute prediction point, and the output is the attribute prediction value of the current point. This module uses the weighted average of the reconstructed attribute values of these three attribute prediction points as the attribute prediction value of the current point: the weights (w_i)_i=1,2,3of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of the three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value Â of the current point,

$Â = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) Attribute residual decoding module: the input of this module is the bitstream and the output is the attribute residual, and this module decodes the attribute residual value of the current point from the bitstream R;
- 4) Reconstructed attribute value acquisition module: the input of the module is attribute residual and attribute prediction value, and the output is the reconstructed attribute value of the current point, and the module obtains the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value: the reconstructed attribute value of the current point is A=Â+R.

Embodiment 9

A point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules.

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the i^thbit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the i^thbit and j^thbit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4;

- Process the attribute values of points in the point cloud in sequence according to Hilbert order, and execute the following modules to process the attribute values of the current points:
- 1) Attribute prediction point acquisition module: the input of the module is the reconstructed point cloud, and the output is the attribute prediction point of the current point, and the module selects the coplanar point and collinear point of the current point among L points in the preceding of Hilbert order as the attribute prediction points of the current point, where the coplanar point is the point whose cube has a common surface with the current point, the collinear point is the point whose cube has a common edge with the current point, and L is a natural number greater than 0, in this embodiment, L is taken as 128: search the first 128 points of the current point in Hilbert order for coplanar and collinear points with the current point k, if k is not 0, let these neighbors be (P_i)_{i=1,2 . . . k}, the attribute value of each neighbor point is (A_i)_{i=1,2 . . . k};
- 2) Attribute prediction point weighting module: the input of this module is attribute prediction point, and the output is attribute prediction point with weight, and this module gives coplanar points and collinear points ω_pland ω_porespectively, in which ω_pland ω_poare real number greater than or equal to 0, ω_plis greater than ω_po, In this embodiment, let ω_pl=2, ω_po=1, the weight of each neighbor point is (ω_i)_{i=1,2 . . . k}, in which the neighbor points with a distance of 1 are given a weight of 2, and the neighbor points with a distance of √{square root over (2)} are given a weight of 2;
- 3) Attribute prediction value calculation module: the input of this module is the weighted attribute prediction point, and the output is the attribute prediction value of the current point. This module uses the weighted average of the reconstructed attribute values of coplanar points and collinear points as the attribute prediction value of the current point: calculate the weighted average of the attribute values of these neighbor points (P_i)_{i=1,2 . . . k}by using the following formula, and take the weighted average value as the attribute prediction value of the current point Â,

$Â = \frac{\sum_{i = 1}^{k} ω_{i} A_{i}}{\sum_{i = 1}^{k} ω_{i}}$

- 4) Attribute residual decoding module: the input of this module is the bitstream and the output is the attribute residual, and this module decodes the attribute residual value of the current point from the bitstream R;
- Reconstructed attribute value acquisition module: the input of the module is attribute residual and attribute prediction value, and the output is the reconstructed attribute value of the current point, and the module obtains the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value: the reconstructed attribute value of the current point is A=Â+R.

Embodiment 10

A point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules.

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

C
_n
=DG((x₁ⁿx₂ⁿx₃ⁿ)₂) 1.

Set X
_i=(x_iⁿ⁺¹. . . x_i^N−1x_i^N),1≤i≤3 2.

Mirror Flip2:if the i^thbit of t₂^Cⁿis 1,makeX_i=2^N−n+1−1−X_i,1≤i≤3 3.

Mirror Flip1:if the i^thbit and j^thbit of T₁^Cⁿis 1,X_i↔X_j,1≤i,j≤3 4;

- LOD building module: The input of this module is the reconstructed point cloud geometry, and the output is the built LOD, this module builds L-level detail level (LOD_i)_{i=0,1 . . . 1−1}, in which, LOD_iis a subset of the reconstructed point cloud, LOD_i⊂LOD_j(i<j), LOD_L−1is the whole reconstructed point cloud, L is a natural number greater than 0, and L is taken as 10 in this embodiment;
- Attribute prediction point acquisition module: the input of the module is the constructed LOD, and the output is the attribute prediction point of the current point, the module acquires n attribute prediction points of the current point, and make (R_i=LOD_i−LOD_i−1)_{i=1 . . . L−1}, R₀=LOD₀, for a point P of Hilbert code H in R_i, n points which are closest to the point P are selected from the points of Hilbert code in the range of [H₁, H₂](H₁≤H≤H₂) in LOD_i−1, and these n points are used as the attribute prediction points of the current point, where N is a natural number, in this embodiment, n is taken as 3. Make these three attribute prediction points as P_i=(X_i, Y_i, Z_i), i=1, 2, 3, set their property values as (A_i)_i=1,2,3;
- Attribute prediction value calculation module: the input of this module is the attribute prediction point, and the output is the attribute prediction value of the current point. This module uses the weighted average of the reconstructed attribute values from these three attribute prediction points is used as the attribute prediction value of the current point: the weights of the three points (ω_i)_{i=1, 2, 3}are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate three points P₁, P₂, P₃. The weighted average of the attribute values of, and take the weighted average as the attribute prediction value of the current point Â,

$Â = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- Attribute residual decoding module: the input of this module is the bitstream and the output is the attribute residual, and this module decodes the attribute residual value of the current point from the bitstream R;
- Reconstructed attribute value acquisition module: the input of the module is attribute residual and attribute prediction value, and the output is the reconstructed attribute value of the current point, and the module obtains the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value: the reconstructed attribute value of the current point is A=Â+R;

Embodiment 11

A point cloud attribute decoding method based on a new processing order in this embodiment, as shown in FIG. 2, includes the following contents,

When the decoded geometric points are in Hilbert order, the point cloud attributes are directly processed according to the decoding order, and the attribute values of the current points are processed by the following steps:

- 1) The weighted average of reconstructed attribute values of preceding decoding points under Hilbert order is used as the attribute prediction value of the current point, where Hilbert order is the point order obtained by sorting points in the point cloud from small to large or from large to small according to their corresponding Hilbert code values: the previous point P of the current point under Hilbert order is selected, set its property values as Â, the property value Â is the predicted value of the attribute of the current point;
- 2) Decoding the attribute residual value of the current point from the bitstream R;
- 3) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 12

A point cloud attribute decoding method based on a new processing order in this embodiment includes the following contents.

- 1) N decoding points in that preceding under Hilbert order are selected as the attribute prediction points of the current point, where n is a natural number, and in this embodiment, n is taken as 3: the first three points P_i=(X_i, Y_i, Z_i), i=1, 2, 3 of the current point under Hilbert order are selected, set their property values as follows (A_i)_i=1,2,3;
- 2) The weighted average of the reconstructed attribute values from these three attribute prediction points is used as the attribute prediction value of the current point: the weights (ω_i)_i=1,2,3of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of the three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value Â of the current point,

$Â = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) Decoding the attribute residual value of the current point from the bitstream R;
- 4) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 13

A point cloud attribute decoding method based on a new processing order in this embodiment includes the following contents.

- 1) Select n points closest to the current point from L points in the preceding of Hilbert order as the attribute prediction points of the current point, where n is a natural number. In this embodiment, n is taken as 3: find 3 points closest to the current point from the first 128 points of the current point in Hilbert order, and make P_i=(X₁, Y_i, Z_i), i=1, 2, 3, set their property values as follows (A_i)_i=1,2,3;
- 2) The weighted average of the reconstructed attribute values from these three attribute prediction points is used as the attribute prediction value of the current point: the weights (ω_i)_i=1, 2, 3 of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of the three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value Â of the current point,

$Â = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 1) Decoding the attribute residual value of the current point from the bitstream R;
- 2) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 14

A point cloud attribute decoding method based on a new processing order in this embodiment includes the following contents.

- 1) Select the coplanar point and collinear point of the current point among L points in the preceding of Hilbert order as the attribute prediction points of the current point, where the coplanar point is the point where the cube where the point is located and the cube where the current point is located have a common surface, and the collinear point refers to the point where the cube where the point is located and the cube where the current point is located have a common edge, and L is a natural number greater than 0. In this embodiment, L is taken as 128: search the first 128 points of the current point in Hilbert order for coplanar and collinear neighbor points of the current point, let the number of these neighbors is k, if k is not 0, let these neighbors be (P_i)_{i=1,2 . . . k}, the attribute value of each neighbor point is (A_i)_{i=1,2 . . . k};
- 2) Weight coplanar points and collinear points ω_pland ω_porespectively, in which ω_pland ω_poare real number greater than or equal to 0, ω_plis greater than ω_po, In this embodiment, let ω_pl=2, ω_po=1, the weight of each neighbor point is (ω_i)_{i=1,2 . . . k}, in which the neighbor points with a distance of 1 are given a weight of 2, and the neighbor points with a distance of √{square root over (2)} are given a weight of 2;
- 3) Use the weighted average of the reconstructed attribute values of coplanar points and collinear points as the attribute prediction value of the current point: calculate the weighted average of the attribute values of these neighbor points (P_i)_{i=1,2 . . . k}by using the following formula, and take the weighted average as the attribute prediction value of the current point Â,

$Â = \frac{\sum_{i = 1}^{k} ω_{i} A_{i}}{\sum_{i = 1}^{k} ω_{i}}$

- 4) Decoding the attribute residual value of the current point from the bitstream R;
- 5) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 15

As shown in FIG. 3, a point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules.

When the decoded geometric points are in Hilbert order, the point cloud attributes are directly processed according to the decoding order, and the following modules are executed for processing the attribute values of the current points:

- 1) Attribute prediction value acquisition module: the input of the module is the reconstructed point cloud, and the output is the attribute prediction value of the current point, and the module uses the weighted average of the reconstructed attribute values of the preceding decoding points in Hilbert order as the attribute prediction value of the current point, wherein the Hilbert order is the point order obtained by sorting the points in the point cloud from small to large or from large to small according to their corresponding Hilbert code values: select the first point of the current point in Hilbert order. PSet their property values as follows Â, the property value Â is the predicted value of the attribute of the current point;
- 2) Attribute residual decoding module: the input of this module is the bitstream and the output is the attribute residual, and this module decodes the attribute residual value of the current point from the bitstream R;
- 3) Reconstructed attribute value acquisition module: the input of the module is attribute residual and attribute prediction value, and the output is the reconstructed attribute value of the current point, and the module obtains the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value: the reconstructed attribute value of the current point is A=Â+R.

Embodiment 16

A point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules.

- 1) Attribute prediction point acquisition module: the input of this module is the reconstructed point cloud, and the output is the attribute prediction point of the current point. This module selects n decoded points in Hilbert order as the attribute prediction points of the current point, where n is a natural number. In this embodiment, n is taken as 3: the first three points P_i=(X_i, Y_i, Z_i), i=1, 2, 3 of the current point in Hilbert order are selected, set their property values as follows (A_i)_{i=1, 2, 3};
- 2) Attribute prediction value calculation module: the input of this module is the attribute prediction point, and the output is the attribute prediction value of the current point. This module uses the weighted average of the reconstructed attribute values of these three attribute prediction points as the attribute prediction value of the current point: the weights (ω_i)_{i=1, 2, 3}of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of the three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value Â of the current point,

$Â = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) Decoding the attribute residual value of the current point from the bitstream R;
- 4) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 17

A point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules.

- 1) Attribute prediction point acquisition module: the input of this module is the reconstructed point cloud, and the output is the attribute prediction point of the current point. This module selects n decoded points in Hilbert order as the attribute prediction points of the current point, where n is a natural number. In this embodiment, n is taken as 3: the first three points P_i=(X₁, Y_i, Z_i), i=1, 2, 3 of the current point in Hilbert order are selected, set their property values as follows (A_i)_i=1,2,3;
- 2) Attribute prediction value calculation module: the input of this module is the attribute prediction point, and the output is the attribute prediction value of the current point. This module uses the weighted average of the reconstructed attribute values of these three attribute prediction points as the attribute prediction value of the current point: the weights (ω_i) i=1, 2, 3 of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of the three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value Â of the current point,

$Â = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) Decoding the attribute residual value of the current point from the bitstream R;
- 4) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 18

A point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules.

- 1) Attribute prediction point acquisition module: the input of the module is the reconstructed point cloud, and the output is the attribute prediction point of the current point, this module selects the coplanar point and collinear point of the current point among L points in the preceding of Hilbert order as the attribute prediction points of the current point, where the coplanar point is the point where the cube where the point is located and the cube where the current point is located have a common surface, and the collinear point refers to the point where the cube where the point is located and the cube where the current point is located have a common edge, and L is a natural number greater than 0. In this embodiment, L is taken as 128: search the first 128 points of the current point in Hilbert order for coplanar and collinear neighbor points of the current point, let the number of these neighbors is k, if k is not 0, let these neighbors be (P_i)_{i=1,2 . . . k}, the attribute value of each neighbor point is (A_i)_{i=1,2 . . . k};
- 2) Attribute prediction point weighting module: the input of the module is the attribute prediction point, and the output is the weighted attribute prediction point, this module weights coplanar points and collinear points ω_pland ω_porespectively, in which ω_pland ω_poare real number greater than or equal to 0, ω_plis greater than ω_po, In this embodiment, let ω_pl=2, ω_po=1, the weight of each neighbor point is (ω_i)_{i=1,2 . . . k}, in which the neighbor points with a distance of 1 are given a weight of 2, and the neighbor points with a distance of √{square root over (2)} are given a weight of 2;
- 3) Attribute prediction value calculation module: The input of this module is a weighted attribute prediction point, and the output is the attribute prediction value of the current point, this module uses the weighted average of the reconstructed attribute values of coplanar points and collinear points as the attribute prediction value of the current point: calculate the weighted average of the attribute values of these neighbor points (P_i)_{i=1,2 . . . k}by using the following formula, and take the weighted average as the attribute prediction value of the current point Â,

$Â = \frac{\sum_{i = 1}^{k} ω_{i} A_{i}}{\sum_{i = 1}^{k} ω_{i}}$

- 4) Decoding the attribute residual value of the current point from the bitstream R;
- 5) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 19

A point cloud attribute decoding method based on a new processing order in this embodiment includes the following contents.

The three-dimensional geometric coordinates (X₁, X₂, X₃)_Nof reconstructed points in the point cloud are processed sequentially from high position to low position of the binary bits, where the i^thprocessing: generate the corresponding Hilbert code subcode C_i,m_iaccording to the (3*m_i) binary bits of the three-dimensional geometric coordinates of the reconstructed point, in this embodiment, N=bitWidth, X₁=x, X₂=y, X₃=z, each fixed only deal with 1 bit, that is, m_i=1, p=bitWidth, ConvertCode^mⁱcorresponds to the table lookup operation HilbertTable [state][pos][1], ConvertState^mⁱcorresponds to the table lookup operation HilbertTable [state][pos][0], the Hilbert code HCode of the decoded point is obtained by the following method:

HCode= 0;

state = 4;

for (i = bitWidth−1; i >= 0; i=i−1) {

HCode= HCode<<3;

bx = (x >> i) & 1;

by = (y >> i) & 1;

bz = (z >> i) & 1;

pos = (bx << 2) | (by << 1) | bz;

HCode= HCode+ HilbertTable [state][pos][1];

state = HilbertTable [state][pos][0];

}

Where >> refers to bitwise right shift operation and << refers to bitwise left shift operation, and the lookup table HilbertCode used is as follows:

HilbertTable[12][8][2] = {

{{5, 0}, {1, 7}, {4, 1}, {2, 6}, {3, 3}, {3, 4}, {4, 2}, {2, 5}},

{{6, 4}, {2, 7}, {6, 3}, {8, 0}, {0, 5}, {0, 6}, {7, 2}, {7, 1}},

{{1, 6}, {0, 7}, {1, 5}, {9, 4}, {10, 1}, {11, 0}, {10, 2}, {9, 3}},

{{9, 2}, {8, 5}, {0, 3}, {0, 4}, {9, 1}, {8, 6}, {6, 0}, {10, 7}},

{{0, 0}, {5, 1}, {8, 3}, {5, 2}, {11, 7}, {6, 6}, {8, 4}, {6, 5}},

{{4, 0}, {10, 3}, {9, 7}, {10, 4}, {0, 1}, {0, 2}, {7, 6}, {7, 5}},

{{11, 6}, {11, 5}, {3, 1}, {3, 2}, {4, 7}, {1, 4}, {9, 0}, {1, 3}},

{{9, 6}, {8, 1}, {5, 7}, {1, 0}, {9, 5}, {8, 2}, {11, 4}, {11, 3}},

{{1, 2}, {4, 3}, {1, 1}, {7, 0}, {10, 5}, {4, 4}, {10, 6}, {3, 7}},

{{2, 4}, {5, 5}, {7, 7}, {5, 6}, {2, 3}, {6, 2}, {3, 0}, {6, 1}},

{{11, 2}, {11, 1}, {3, 5}, {3, 6}, {5, 3}, {2, 0}, {5, 4}, {8, 7}},

{{7, 4}, {7, 3}, {4, 5}, {2, 2}, {6, 7}, {10, 0}, {4, 6}, {2, 1}},

According to the sequence of Hilbert from small to large or from large to small, the reconstructed attribute information of the point cloud is decoded from the bitstream in order, and the attribute value of the current point is processed by the following steps:

- 1) Select n points closest to the current point from L points in the preceding of Hilbert order as the attribute prediction points of the current point, where n is a natural number. In this embodiment, n is taken as 3: find 3 points closest to the current point from the first 128 points of the current point in Hilbert order, and make P_i=(X₁, Y_i, Z_i), i=1, 2, 3, set their property values as follows(A_i)_{i=1, 2, 3};
- 2) The weighted average of the reconstructed attribute values from these three attribute prediction points is used as the attribute prediction value of the current point: the weights (ω_i)_i=1,2,3 of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}},$

$i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of the three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value A of the current point,

$\hat{A} = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) Decoding the attribute residual value of the current point from the bitstream R;
- 4) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 20

A point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules.

Hilbert code generation module: the input of this module is the reconstructed geometric coordinates of the point cloud, and the output is the Hilbert code corresponding to the geometric coordinates of the point cloud. This module sequentially processes the three-dimensional geometric coordinates (X₁, X₂, X₃)_Nof reconstructed points in the point cloud sequentially from high position to low position of the binary bits, where the i^thprocessing: generate the corresponding Hilbert code subcode C_i,m_iaccording to the (3*m_i) binary bits of the three-dimensional geometric coordinates of the reconstructed point, in this embodiment, N=bitWidth, X_i=x, X₂=y, X₃=z, each fixed only deal with 1 bit, that is, m_i=1, p=bitWidth, ConvertCode^mⁱcorresponds to the table lookup operation HilbertTable [state][pos][1], ConvertState^mⁱcorresponds to the table lookup operation HilbertTable [state][pos][0], the Hilbert code HCode of the decoded point is obtained by the following method:

HCode= 0;

state = 4;

for (i = bitWidth−1; i >= 0; i=i−1) {

HCode= HCode<<3;

bx = (x >> i) & 1;

by = (y >> i) & 1;

bz = (z >> i) & 1;

pos = (bx << 2) | (by << 1) | bz;

HCode = HCode+ Hilbert Table [state ][pos][1];

state = HilbertTable [state ][pos][0];

}

- Where >> refers to bitwise right shift operation and << refers to bitwise left shift operation, and the lookup table of HilbertCode used is as follows:

Attribute decoding module: the input of this module is the attribute bitstream, and the output is the reconstructed point cloud attribute. This module decodes the reconstructed point cloud attribute information from the bitstream in order of Hilbert from small to large or from large to small, and executes the following modules to process the attribute value of the current point:

- 1) Attribute prediction point acquisition module: the input of this module is the reconstructed point cloud, and the output is the attribute prediction point of the current point. The module selects n points closest to the current point from the L points in the preceding under Hilbert order as the attribute prediction points of the current point, where n is a natural number, and in this embodiment, n is taken as 3: three points closest to the current point are found in the first 128 points of the current point under Hilbert order, and make these as P_i=(X₁, i=1, 2, 3, set their property values as follows (A_i)_{i=1, 2, 3};
- 2) Attribute prediction value calculation module: the input of this module is the attribute prediction point, and the output is the attribute prediction value of the current point. This module uses the weighted average of the reconstructed attribute values of these three attribute prediction points as the attribute prediction value of the current point: the weights (ω_i)_{i=1, 2, 3}of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}}, i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are weighting coefficients of three dimensions, which are used to control the weights of the three dimensions. In this embodiment, the weighting coefficient a, b, c are respectively valued as a=2, b=1, c=3;
- Use the following formula to calculate the weighted average of the attribute values of the three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value Â of the current point,

$\hat{A} = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) Decoding the attribute residual value of the current point from the bitstream R;
- 4) Obtaining the reconstructed attribute value of the current point according to the attribute prediction value of the current point and the decoded attribute residual value, wherein the reconstructed attribute value of the current point is A=Â+R.

Embodiment 21

A point cloud attribute decoding method based on a new processing order in this embodiment includes the following contents.

The three-dimensional geometric coordinates (X₁, X₂, X₃)_Nof reconstructed points in the point cloud are processed sequentially from high position to low position of the binary bits, where the i^thprocessing: generate the corresponding Hilbert code subcode C_i,m_iaccording to the (3*m_i) binary bits of the three-dimensional geometric coordinates of the reconstructed point, in this embodiment, N=bitWidth and even, X₁=x, X₂=y, X₃=z, each fixed only deal with 2 bit, that is, m_i=2, p=bitWidth/2, ConvertCode^mⁱcorresponds to the table lookup operation HilbertTable [state][pos][1], ConvertState^mⁱcorresponds to the table lookup operation HilbertTable [state][pos][1], the Hilbert code HCode of the decoded point is obtained by the following method:

HCode= 0;

state = 4;

for (i = bitWidth−2; i >= 0; i−=2) {

HilbertCode <<= 6;

bx = (x >> i) & 3;

by = (y >> 1) & 3;

bz = (z >> i) & 3;

pos = (bx << 4) | (by << 2) | bz;

HilbertCode += HilbertTable [state][pos][1];

state = HilbertTable [state][pos][0];

}

Where >> refers to bitwise right shift operation and << refers to bitwise left shift operation, and the lookup table of HilbertCode used is as follows:

HilbertTable [12][64][2] = {

{{4, 0}, {10, 3}, {6, 60}, {2, 63}, {9, 7}, {10, 4}, {6, 59}, {8, 56}, {0, 8}, {5, 9}, {1, 54}, {0, 55}, {8, 11},

{5, 10}, {1, 53}, {9, 52}, {0, 1}, {0, 2}, {0, 61}, {0, 62}, {7, 6}, {7, 5}, {7, 58}, {7, 57}, {11, 15}, {6, 14}, {10, 49},

{11, 48}, {8, 12}, {6, 13}, {10, 50}, {9, 51}, {9, 26}, {8, 29}, {9, 34}, {8, 37}, {0, 27}, {0, 28}, {0, 35}, {0, 36},

{0, 16}, {5, 17}, {1, 46}, {0, 47}, {8, 19}, {5, 18}, {1, 45}, {9, 44}, {9, 25}, {8, 30}, {9, 33}, {8, 38}, {6, 24},

{10, 31}, {6, 32}, {10, 39}, {11, 23}, {6, 22}, {10, 41}, {11, 40}, {8, 20}, {6, 21}, {10, 42}, {9, 43}},

{{11, 38}, {11, 37}, {1, 62}, {0, 63}, {3, 33}, {3, 34}, {1, 61}, {9, 60}, {11, 30}, {11, 29}, {1, 2}, {4, 3}, {3, 25},

{3, 26}, {1, 1}, {7, 0}, {4, 39}, {1, 36}, {10, 57}, {11, 56}, {9, 32}, {1, 35}, {10, 58}, {9, 59}, {4, 31}, {1, 28},

{10, 5}, {4, 4}, {9, 24}, {1, 27}, {10, 6}, {3, 7}, {5, 40}, {1, 47}, {5, 48}, {1, 55}, {4, 41}, {2, 46}, {4, 49},

{2, 54}, {9, 22}, {8, 17}, {9, 14}, {8, 9}, {5, 23}, {1, 16}, {5, 15}, {1, 8}, {3, 43}, {3, 44}, {3, 51}, {3, 52},

{4, 42}, {2, 45}, {4, 50}, {2, 53}, {9, 21}, {8, 18}, {9, 13}, {8, 10}, {11, 20}, {11, 19}, {11, 12}, {11, 11}},

{{6, 52}, {2, 55}, {5, 56}, {1, 63}, {6, 51}, {8, 48}, {4, 57}, {2, 62}, {6, 44}, {2, 47}, {2, 36}, {5, 37}, {6, 43},

{8, 40}, {7, 39}, {5, 38}, {0, 53}, {0, 54}, {3, 59}, {3, 60}, {7, 50}, {7, 49}, {4, 58}, {2, 61}, {0, 45}, {0, 46},

{2, 35}, {6, 34}, {7, 42}, {7, 41}, {3, 32}, {6, 33}, {11, 10}, {11, 9}, {7, 4}, {7, 3}, {3, 13}, {3, 14}, {4, 5}, {2, 2},

{11, 18}, {11, 17}, {2, 28}, {5, 29}, {3, 21}, {3, 22}, {7, 31}, {5, 30}, {5, 11}, {2, 8}, {6, 7}, {10, 0}, {5, 12},

{8, 15}, {4, 6}, {2, 1}, {5, 19}, {2, 16}, {2, 27}, {6, 26}, {5, 20}, {8, 23}, {3, 24}, {6, 25}},

{{2, 20}, {5, 21}, {1, 42}, {4, 43}, {7, 23}, {5, 22}, {1, 41}, {7, 40}, {5, 24}, {1, 31}, {5, 32}, {1, 39}, {4, 25},

{2, 30}, {4, 33}, {2, 38}, {2, 19}, {6, 18}, {10, 45}, {4, 44}, {3, 16}, {6, 17}, {10, 46}, {3, 47}, {3, 27}, {3, 28},

{3, 35}, {3, 36}, {4, 26}, {2, 29}, {4, 34}, {2, 37}, {2, 12}, {5, 13}, {1, 50}, {4, 51}, {7, 15}, {5, 14}, {1, 49},

{7, 48}, {11, 6}, {11, 5}, {11, 58}, {11, 57}, {3, 1}, {3, 2}, {3, 61}, {3, 62}, {2, 11}, {6, 10}, {10, 53}, {4, 52},

{3, 8}, {6, 9}, {10, 54}, {3, 55}, {4, 7}, {1, 4}, {5, 59}, {2, 56}, {9, 0}, {1, 3}, {5, 60}, {8, 63}},

{{5, 0}, {1, 7}, {4, 8}, {10, 11}, {4, 1}, {2, 6}, {9, 15}, {10, 12}, {1, 26}, {4, 27}, {4, 16}, {10, 19}, {1, 25},

{7, 24}, {9, 23}, {10, 20}, {3, 3}, {3, 4}, {0, 9}, {0, 10}, {4, 2}, {2, 5}, {7, 14}, {7, 13}, {10, 29}, {4, 28}, {0, 17},

{0, 18}, {10, 30}, {3, 31}, {7, 22}, {7, 21}, {7, 60}, {7, 59}, {11, 54}, {11, 53}, {4, 61}, {2, 58}, {3, 49}, {3, 50},

{1, 34}, {4, 35}, {11, 46}, {11, 45}, {1, 33}, {7, 32}, {3, 41}, {3, 42}, {6, 63}, {10, 56}, {4, 55}, {1, 52}, {4, 62},

{2, 57}, {9, 48}, {1, 51}, {10, 37}, {4, 36}, {4, 47}, {1, 44}, {10, 38}, {3, 39}, {9, 40}, {1, 43}},

{{0, 0}, {5, 1}, {11, 26}, {11, 25}, {8, 3}, {5, 2}, {3, 29}, {3, 30}, {2, 60}, {5, 61}, {11, 34}, {11, 33}, {7, 63},

{5, 62}, {3, 37}, {3, 38}, {11, 7}, {6, 6}, {5, 27}, {2, 24}, {8, 4}, {6, 5}, {5, 28}, {8, 31}, {2, 59}, {6, 58},

{5, 35}, {2, 32}, {3, 56}, {6, 57}, {5, 36}, {8, 39}, {5, 8}, {1, 15}, {5, 16}, {1, 23}, {4, 9}, {2, 14}, {4, 17}, {2, 22},

{9, 54}, {8, 49}, {9, 46}, {8, 41}, {5, 55}, {1, 48}, {5, 47}, {1, 40}, {3, 11}, {3, 12}, {3, 19}, {3, 20}, {4, 10},

{2, 13}, {4, 18}, {2, 21}, {9, 53}, {8, 50}, {9, 45}, {8, 42}, {11, 52}, {11, 51}, {11, 44}, {11, 43}},

{{7, 52}, {7, 51}, {7, 44}, {7, 43}, {4, 53}, {2, 50}, {4, 45}, {2, 42}, {9, 10}, {8, 13}, {9, 18}, {8, 21}, {0, 11},

{0, 12}, {0, 19}, {0, 20}, {6, 55}, {10, 48}, {6, 47}, {10, 40}, {4, 54}, {2, 49}, {4, 46}, {2, 41}, {9, 9}, {8, 14},

{9, 17}, {8, 22}, {6, 8}, {10, 15}, {6, 16}, {10, 23}, {0, 56}, {5, 57}, {6, 36}, {2, 39}, {8, 59}, {5, 58}, {6, 35},

{8, 32}, {2, 4}, {5, 5}, {6, 28}, {2, 31}, {7, 7}, {5, 6}, {6, 27}, {8, 24}, {11, 63}, {6, 62}, {0, 37}, {0, 38},

{8, 60}, {6, 61}, {7, 34}, {7, 33}, {2, 3}, {6, 2}, {0, 29}, {0, 30}, {3, 0}, {6, 1}, {7, 26}, {7, 25}},

{{2, 52}, {5, 53}, {1, 10}, {4, 11}, {7, 55}, {5, 54}, {1, 9}, {7, 8}, {4, 56}, {10, 59}, {6, 4}, {2, 7}, {9, 63},

{10, 60}, {6, 3}, {8, 0}, {2, 51}, {6, 50}, {10, 13}, {4, 12}, {3, 48}, {6, 49}, {10, 14}, {3, 15}, {0, 57}, {0, 58},

{0, 5}, {0, 6}, {7, 62}, {7, 61}, {7, 2}, {7, 1}, {2, 44}, {5, 45}, {1, 18}, {4, 19}, {7, 47}, {5, 46}, {1, 17}, {7, 16},

{7, 36}, {7, 35}, {7, 28}, {7, 27}, {4, 37}, {2, 34}, {4, 29}, {2, 26}, {2, 43}, {6, 42}, {10, 21}, {4, 20}, {3, 40},

{6, 41}, {10, 22}, {3, 23}, {6, 39}, {10, 32}, {6, 31}, {10, 24}, {4, 38}, {2, 33}, {4, 30}, {2, 25}},

{{6, 20}, {2, 23}, {0, 24}, {5, 25}, {6, 19}, {8, 16}, {8, 27}, {5, 26}, {6, 12}, {2, 15}, {9, 6}, {8, 1}, {6, 11},

{8, 8}, {5, 7}, {1, 0}, {0, 21}, {0, 22}, {11, 31}, {6, 30}, {7, 18}, {7, 17}, {8, 28}, {6, 29}, {0, 13}, {0, 14}, {9, 5},

{8, 2}, {7, 10}, {7, 9}, {11, 4}, {11, 3}, {11, 42}, {11, 41}, {0, 32}, {5, 33}, {3, 45}, {3, 46}, {8, 35}, {5, 34},

{11, 50}, {11, 49}, {9, 58}, {8, 61}, {3, 53}, {3, 54}, {0, 59}, {0, 60}, {5, 43}, {2, 40}, {11, 39}, {6, 38}, {5, 44},

{8, 47}, {8, 36}, {6, 37}, {5, 51}, {2, 48}, {9, 57}, {8, 62}, {5, 52}, {8, 55}, {6, 56}, {10, 63}},

{{1, 38}, {0, 39}, {4, 40}, {10, 43}, {1, 37}, {9, 36}, {9, 47}, {10, 44}, {9, 62}, {8, 57}, {4, 48}, {10, 51}, {5, 63},

{1, 56}, {9, 55}, {10, 52}, {10, 33}, {11, 32}, {0, 41}, {0, 42}, {10, 34}, {9, 35}, {7, 46}, {7, 45}, {9, 61}, {8, 58},

{0, 49}, {0, 50}, {11, 60}, {11, 59}, {7, 54}, {7, 53}, {1, 30}, {0, 31}, {11, 22}, {11, 21}, {1, 29}, {9, 28}, {3, 17},

{3, 18}, {9, 2}, {8, 5}, {11, 14}, {11, 13}, {0, 3}, {0, 4}, {3, 9}, {3, 10}, {10, 25}, {11, 24}, {4, 23}, {1, 20},

{10, 26}, {9, 27}, {9, 16}, {1, 19}, {9, 1}, {8, 6}, {4, 15}, {1, 12}, {6, 0}, {10, 7}, {9, 8}, {1, 11}},

{{7, 20}, {7, 19}, {7, 12}, {7, 11}, {4, 21}, {2, 18}, {4, 13}, {2, 10}, {9, 42}, {8, 45}, {9, 50}, {8, 53}, {0, 43},

{0, 44}, {0, 51}, {0, 52}, {6, 23}, {10, 16}, {6, 15}, {10, 8}, {4, 22}, {2, 17}, {4, 14}, {2, 9}, {9, 41}, {8, 46},

{9, 49}, {8, 54}, {6, 40}, {10, 47}, {6, 48}, {10, 55}, {4, 24}, {10, 27}, {1, 6}, {0, 7}, {9, 31}, {10, 28}, {1, 5},

{9, 4}, {4, 32}, {10, 35}, {1, 58}, {4, 59}, {9, 39}, {10, 36}, {1, 57}, {7, 56}, {0, 25}, {0, 26}, {10, 1}, {11, 0},

{7, 30}, {7, 29}, {10, 2}, {9, 3}, {0, 33}, {0, 34}, {10, 61}, {4, 60}, {7, 38}, {7, 37}, {10, 62}, {3, 63}},

{{9, 38}, {8, 33}, {9, 30}, {8, 25}, {5, 39}, {1, 32}, {5, 31}, {1, 24}, {0, 40}, {5, 41}, {1, 22}, {0, 23}, {8, 43},

{5, 42}, {1, 21}, {9, 20}, {9, 37}, {8, 34}, {9, 29}, {8, 26}, {11, 36}, {11, 35}, {11, 28}, {11, 27}, {11, 47},

{6, 46}, {10, 17}, {11, 16}, {8, 44}, {6, 45}, {10, 18}, {9, 19}, {11, 62}, {11, 61}, {11, 2}, {11, 1}, {3, 57}, {3, 58},

{3, 5}, {3, 6}, {0, 48}, {5, 49}, {1, 14}, {0, 15}, {8, 51}, {5, 50}, {1, 13}, {9, 12}, {4, 63}, {1, 60}, {5, 3},

{2, 0}, {9, 56}, {1, 59}, {5, 4}, {8, 7}, {11, 55}, {6, 54}, {10, 9}, {11, 8}, {8, 52}, {6, 53}, {10, 10}, {9, 11}}};

According to the Hilbert order from small to large or from large to small, the reconstructed attribute information of the point cloud is decoded from the bitstream in order, and the attribute value of the current point is processed by the following steps:

- 1) Select the nearest point of n among the L points in the preceding sequence of Hilbert order as the attribute prediction point of the current point, n is a natural number, and in this example, n is 3: find three points closest to the current point among the first 128 points of the current point under Hilbert order, and set them as P_i=(X₁, Y_i, Z_i), i=1, 2, 3, and set their attribute values as (A_i)_{i=1, 2, 3};
- 2) Use the weighted average value of the three attribute prediction points to reconstruct the attribute values as the attribute prediction value of the current point: the weight (ω_i)_i=1,2,3of the three points are obtained respectively.

$ω_{i} = \frac{1}{d_{i}}, i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are the weighting coefficients of the three dimensions, which are used to control the weight of the three dimensions. In this embodiment, the weighting coefficients a, b, c are a=2, b=1 and c=3 respectively.

Use the following formula to calculate the weighted average of the attribute values of the three points P₁, P₂, P₃, and take the weighted average value as the attribute prediction value Â of the current point,

$\hat{A} = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) The attribute residual value R of the current point is decoded from the bitstream;
- 4) According to the attribute prediction value of the current point and the decoded attribute residual value, the reconstructed attribute value of the current point is A=Â+R.

Embodiment 22

A point cloud attribute decoding device based on a new processing order in this embodiment includes the following modules.

The binary bits of the three-dimensional geometric coordinates (X₁, X₂, X₃)_Nof the reconstructed points in the point cloud are processed in order from high to low, and the i th processing: According to the (3*m_i) binary bits in the three-dimensional geometric coordinates of the reconstructed points, the corresponding Hilbert code is generated. In this embodiment, N=bitWidth and is an even number. X₁=x, X₂=y, X₃=z. Only 2 bits are processed each time, that is, m_i=2, p=bitWidth/2. ConvertCode^mⁱcorresponds to HilbertTable [state] [pos] [1], ConvertState^mⁱcorresponds to HilbertTable [state] [pos] [0]. The Hilbert code HCode of the decoded point is obtained by the following method:

HCode= 0;

state = 4;

for (i = bitWidth−2; i >= 0; i−=2) {

HilbertCode <<= 6;

bx = (x >> i) & 3;

by = (y >> i) & 3;

bz = (z >> i) & 3;

pos = (bx << 4) | (by << 2) | bz;

HilbertCode += HilbertTable [state][pos][1];

state = HilbertTable [state][pos][0];

}

- Where >> refers to bitwise right shift operation and << refers to bitwise left shift operation, and the lookup table HilbertCode used is as follows:

HilbertTable [12][64][2] = {

{{4, 0}, {10, 3}, {6, 60}, {2, 63}, {9, 7}, {10, 4}, {6, 59}, {8, 56}, {0, 8}, {5, 9}, {1, 54}, {0, 55}, {8, 11}, {5, 10},

{1, 53}, {9, 52}, {0, 1}, {0, 2}, {0, 61}, {0, 62}, {7, 6}, {7, 5}, {7, 58}, {7, 57}, {11, 15}, {6, 14}, {10, 49},

{11, 48}, {8, 12}, {6, 13}, {10, 50}, {9, 51}, {9, 26}, {8, 29}, {9, 34}, {8, 37}, {0, 27}, {0, 28}, {0, 35}, {0, 36},

{0, 16}, {5, 17}, {1, 46}, {0, 47}, {8, 19}, {5, 18}, {1, 45}, {9, 44}, {9, 25}, {8, 30}, {9, 33}, {8, 38}, {6, 24},

{10, 31}, {6, 32}, {10, 39}, {11, 23}, {6, 22}, {10, 41}, {11, 40}, {8, 20}, {6, 21}, {10, 42}, {9, 43}},

{{11, 38}, {11, 37}, {1, 62}, {0, 63}, {3, 33}, {3, 34}, {1, 61}, {9, 60}, {11, 30}, {11, 29}, {1, 2}, {4, 3},

{3, 25}, {3, 26}, {1, 1}, {7, 0}, {4, 39}, {1, 36}, {10, 57}, {11, 56}, {9, 32}, {1, 35}, {10, 58}, {9, 59}, {4, 31},

{1, 28}, {10, 5}, {4, 4}, {9, 24}, {1, 27}, {10, 6}, {3, 7}, {5, 40}, {1, 47}, {5, 48}, {1, 55}, {4, 41}, {2, 46}, {4, 49},

{2, 54}, {9, 22}, {8, 17}, {9, 14}, {8, 9}, {5, 23}, {1, 16}, {5, 15}, {1, 8}, {3, 43}, {3, 44}, {3, 51}, {3, 52},

{4, 42}, {2, 45}, {4, 50}, {2, 53}, {9, 21}, {8, 18}, {9, 13}, {8, 10}, {11, 20}, {11, 19}, {11, 12}, {11, 11}},

{{6, 52}, {2, 55}, {5, 56}, {1, 63}, {6, 51}, {8, 48}, {4, 57}, {2, 62}, {6, 44}, {2, 47}, {2, 36}, {5, 37}, {6, 43},

{8, 40}, {7, 39}, {5, 38}, {0, 53}, {0, 54}, {3, 59}, {3, 60}, {7, 50}, {7, 49}, {4, 58}, {2, 61}, {0, 45}, {0, 46},

{2, 35}, {6, 34}, {7, 42}, {7, 41}, {3, 32}, {6, 33}, {11, 10}, {11, 9}, {7, 4}, {7, 3}, {3, 13}, {3, 14}, {4, 5}, {2, 2},

{11, 18}, {11, 17}, {2, 28}, {5, 29}, {3, 21}, {3, 22}, {7, 31}, {5, 30}, {5, 11}, {2, 8}, {6, 7}, {10, 0}, {5, 12},

{8, 15}, {4, 6}, {2, 1}, {5, 19}, {2, 16}, {2, 27}, {6, 26}, {5, 20}, {8, 23}, {3, 24}, {6, 25}},

{{2, 20}, {5, 21}, {1, 42}, {4, 43}, {7, 23}, {5, 22}, {1, 41}, {7, 40}, {5, 24}, {1, 31}, {5, 32}, {1, 39}, {4, 25},

{2, 30}, {4, 33}, {2, 38}, {2, 19}, {6, 18}, {10, 45}, {4, 44}, {3, 16}, {6, 17}, {10, 46}, {3, 47}, {3, 27}, {3, 28},

{3, 35}, {3, 36}, {4, 26}, {2, 29}, {4, 34}, {2, 37}, {2, 12}, {5, 13}, {1, 50}, {4, 51}, {7, 15}, {5, 14}, {1, 49},

{7, 48}, {11, 6}, {11, 5}, {11, 58}, {11, 57}, {3, 1}, {3, 2}, {3, 61}, {3, 62}, {2, 11}, {6, 10}, {10, 53}, {4, 52},

{3, 8}, {6, 9}, {10, 54}, {3, 55}, {4, 7}, {1, 4}, {5, 59}, {2, 56}, {9, 0}, {1, 3}, {5, 60}, {8, 63}},

{{5, 0}, {1, 7}, {4, 8}, {10, 11}, {4, 1}, {2, 6}, {9, 15}, {10, 12}, {1, 26}, {4, 27}, {4, 16}, {10, 19}, {1, 25},

{7, 24}, {9, 23}, {10, 20}, {3, 3}, {3, 4}, {0, 9}, {0, 10}, {4, 2}, {2, 5}, {7, 14}, {7, 13}, {10, 29}, {4, 28}, {0, 17},

{0, 18}, {10, 30}, {3, 31}, {7, 22}, {7, 21}, {7, 60}, {7, 59}, {11, 54}, {11, 53}, {4, 61}, {2, 58}, {3, 49}, {3, 50},

{1, 34}, {4, 35}, {11, 46}, {11, 45}, {1, 33}, {7, 32}, {3, 41}, {3, 42}, {6, 63}, {10, 56}, {4, 55}, {1, 52}, {4, 62},

{2, 57}, {9, 48}, {1, 51}, {10, 37}, {4, 36}, {4, 47}, {1, 44}, {10, 38}, {3, 39}, {9, 40}, {1, 43}},

{{0, 0}, {5, 1}, {11, 26}, {11, 25}, {8, 3}, {5, 2}, {3, 29}, {3, 30}, {2, 60}, {5, 61}, {11, 34}, {11, 33}, {7, 63},

{5, 62}, {3, 37}, {3, 38}, {11, 7}, {6, 6}, {5, 27}, {2, 24}, {8, 4}, {6, 5}, {5, 28}, {8, 31}, {2, 59}, {6, 58}, {5, 35},

{2, 32}, {3, 56}, {6, 57}, {5, 36}, {8, 39}, {5, 8}, {1, 15}, {5, 16}, {1, 23}, {4, 9}, {2, 14}, {4, 17}, {2, 22},

{9, 54}, {8, 49}, {9, 46}, {8, 41}, {5, 55}, {1, 48}, {5, 47}, {1, 40}, {3, 11}, {3, 12}, {3, 19}, {3, 20}, {4, 10},

{2, 13}, {4, 18}, {2, 21}, {9, 53}, {8, 50}, {9, 45}, {8, 42}, {11, 52}, {11, 51}, {11, 44}, {11, 43}},

{{7, 52}, {7, 51}, {7, 44}, {7, 43}, {4, 53}, {2, 50}, {4, 45}, {2, 42}, {9, 10}, {8, 13}, {9, 18}, {8, 21}, {0, 11},

{0, 12}, {0, 19}, {0, 20}, {6, 55}, {10, 48}, {6, 47}, {10, 40}, {4, 54}, {2, 49}, {4, 46}, {2, 41}, {9, 9}, {8, 14},

{9, 17}, {8, 22}, {6, 8}, {10, 15}, {6, 16}, {10, 23}, {0, 56}, {5, 57}, {6, 36}, {2, 39}, {8, 59}, {5, 58}, {6, 35},

{8, 32}, {2, 4}, {5, 5}, {6, 28}, {2, 31}, {7, 7}, {5, 6}, {6, 27}, {8, 24}, {11, 63}, {6, 62}, {0, 37}, {0, 38},

{8, 60}, {6, 61}, {7, 34}, {7, 33}, {2, 3}, {6, 2}, {0, 29}, {0, 30}, {3, 0}, {6, 1}, {7, 26}, {7, 25}},

{{2, 52}, {5, 53}, {1, 10}, {4, 11}, {7, 55}, {5, 54}, {1, 9}, {7, 8}, {4, 56}, {10, 59}, {6, 4}, {2, 7}, {9, 63},

{10, 60}, {6, 3}, {8, 0}, {2, 51}, {6, 50}, {10, 13}, {4, 12}, {3, 48}, {6, 49}, {10, 14}, {3, 15}, {0, 57}, {0, 58},

{0, 5}, {0, 6}, {7, 62}, {7, 61}, {7, 2}, {7, 1}, {2, 44}, {5, 45}, {1, 18}, {4, 19}, {7, 47}, {5, 46}, {1, 17}, {7, 16},

{7, 36}, {7, 35}, {7, 28}, {7, 27}, {4, 37}, {2, 34}, {4, 29}, {2, 26}, {2, 43}, {6, 42}, {10, 21}, {4, 20}, {3, 40},

{6, 41}, {10, 22}, {3, 23}, {6, 39}, {10, 32}, {6, 31}, {10, 24}, {4, 38}, {2, 33}, {4, 30}, {2, 25}},

{{6, 20}, {2, 23}, {0, 24}, {5, 25}, {6, 19}, {8, 16}, {8, 27}, {5, 26}, {6, 12}, {2, 15}, {9, 6}, {8, 1}, {6, 11},

{8, 8}, {5, 7}, {1, 0}, {0, 21}, {0, 22}, {11, 31}, {6, 30}, {7, 18}, {7, 17}, {8, 28}, {6, 29}, {0, 13}, {0, 14}, {9, 5},

{8, 2}, {7, 10}, {7, 9}, {11, 4}, {11, 3}, {11, 42}, {11, 41}, {0, 32}, {5, 33}, {3, 45}, {3, 46}, {8, 35}, {5, 34},

{11, 50}, {11, 49}, {9, 58}, {8, 61}, {3, 53}, {3, 54}, {0, 59}, {0, 60}, {5, 43}, {2, 40}, {11, 39}, {6, 38}, {5, 44},

{8, 47}, {8, 36}, {6, 37}, {5, 51}, {2, 48}, {9, 57}, {8, 62}, {5, 52}, {8, 55}, {6, 56}, {10, 63}},

{{1, 38}, {0, 39}, {4, 40}, {10, 43}, {1, 37}, {9, 36}, {9, 47}, {10, 44}, {9, 62}, {8, 57}, {4, 48}, {10, 51}, {5, 63},

{1, 56}, {9, 55}, {10, 52}, {10, 33}, {11, 32}, {0, 41}, {0, 42}, {10, 34}, {9, 35}, {7, 46}, {7, 45}, {9, 61}, {8, 58},

{0, 49}, {0, 50}, {11, 60}, {11, 59}, {7, 54}, {7, 53}, {1, 30}, {0, 31}, {11, 22}, {11, 21}, {1, 29}, {9, 28}, {3, 17},

{3, 18}, {9, 2}, {8, 5}, {11, 14}, {11, 13}, {0, 3}, {0, 4}, {3, 9}, {3, 10}, {10, 25}, {11, 24}, {4, 23}, {1, 20},

{10, 26}, {9, 27}, {9, 16}, {1, 19}, {9, 1}, {8, 6}, {4, 15}, {1, 12}, {6, 0}, {10, 7}, {9, 8}, {1, 11}},

{{7, 20}, {7, 19}, {7, 12}, {7, 11}, {4, 21}, {2, 18}, {4, 13}, {2, 10}, {9, 42}, {8, 45}, {9, 50}, {8, 53}, {0, 43},

{0, 44}, {0, 51}, {0, 52}, {6, 23}, {10, 16}, {6, 15}, {10, 8}, {4, 22}, {2, 17}, {4, 14}, {2, 9}, {9, 41}, {8, 46},

{9, 49}, {8, 54}, {6, 40}, {10, 47}, {6, 48}, {10, 55}, {4, 24}, {10, 27}, {1, 6}, {0, 7}, {9, 31}, {10, 28}, {1, 5},

{9, 4}, {4, 32}, {10, 35}, {1, 58}, {4, 59}, {9, 39}, {10, 36}, {1, 57}, {7, 56}, {0, 25}, {0, 26}, {10, 1}, {11, 0},

{7, 30}, {7, 29}, {10, 2}, {9, 3}, {0, 33}, {0, 34}, {10, 61}, {4, 60}, {7, 38}, {7, 37}, {10, 62}, {3, 63}},

{{9, 38}, {8, 33}, {9, 30}, {8, 25}, {5, 39}, {1, 32}, {5, 31}, {1, 24}, {0, 40}, {5, 41}, {1, 22}, {0, 23}, {8, 43},

{5, 42}, {1, 21}, {9, 20}, {9, 37}, {8, 34}, {9, 29}, {8, 26}, {11, 36}, {11, 35}, {11, 28}, {11, 27}, {11, 47},

{6, 46}, {10, 17}, {11, 16}, {8, 44}, {6, 45}, {10, 18}, {9, 19}, {11, 62}, {11, 61}, {11, 2}, {11, 1}, {3, 57}, {3, 58},

{3, 5}, {3, 6}, {0, 48}, {5, 49}, {1, 14}, {0, 15}, {8, 51}, {5, 50}, {1, 13}, {9, 12}, {4, 63}, {1, 60}, {5, 3}, {2, 0},

{9, 56}, {1, 59}, {5, 4}, {8, 7}, {11, 55}, {6, 54}, {10, 9}, {11, 8}, {8, 52}, {6, 53}, {10, 10}, {9, 11}}};

{{9, 38}, {8, 33}, {9, 30}, {8, 25}, {5, 39}, {1, 32}, {5, 31}, {1, 24}, {0, 40}, {5, 41}, {1, 22}, {0, 23}, {8, 43},

{5, 42}, {1, 21}, {9, 20}, {9, 37}, {8, 34}, {9, 29}, {8, 26}, {11, 36}, {11, 35}, {11, 28}, {11, 27}, {11, 47},

{6, 46}, {10, 17}, {11, 16}, {8, 44}, {6, 45}, {10, 18}, {9, 19}, {11, 62}, {11, 61}, {11, 2}, {11, 1}, {3, 57}, {3, 58},

{3, 5}, {3, 6}, {0, 48}, {5, 49}, {1, 14}, {0, 15}, {8, 51}, {5, 50}, {1, 13}, {9, 12}, {4, 63}, {1, 60}, {5, 3}, {2, 0},

{9, 56}, {1, 59}, {5, 4}, {8, 7}, {11, 55}, {6, 54}, {10, 9}, {11, 8}, {8, 52}, {6, 53}, {10, 10}, {9, 11}}};

Attribute decoding module: the input of the module is attribute bitstream and the output is reconstructed attribute of the point cloud. According to the Hilbert order from small to large or from large to small, the reconstructed attribute information of the point cloud is decoded from the bitstream in order, and the attribute value of the current point is processed by the following steps:

- 3) Attribute prediction point acquisition module: The input of this module is the reconstructed point cloud, and the output is the attribute prediction point of the current point. The module selects the nearest point of n among the L points in the preceding sequence of Hilbert order as the attribute prediction point of the current point, n is a natural number, and in this example, n is 3: find three points closest to the current point among the first 128 points of the current point under Hilbert order, and set them as P_i=(X_i, Y_i, Z_i) i=1, 2, 3, and set their attribute values as (A_i)_{i=1, 2, 3};
- 4) Attribute prediction value calculation module: the input of the module is the attribute prediction point, and the output is the attribute prediction value of the current point. The module uses the weighted average value of the three attribute prediction points to reconstruct the attribute values as the attribute prediction value of the current point: the weight (ω_i)_i=1,2,3of the three points are obtained respectively,

$ω_{i} = \frac{1}{d_{i}}, i = 1, 2, 3$

Among them, d_iis the distance between the point P_iand the current point (X, Y, Z), which can be at least one of the following distances:

- Euclidean distance: d_i=√{square root over (a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²)}
- Square of Euclidean distance: d_i=a(X_i−X)²+b(Y_i−Y)²+c(Z_i−Z)²
- Chebyshev distance: d_i=max(a|X_i−X|, b|Y_i−Y|, c|Z_i−Z|)
- Manhattan distance: d_i=a|X_i−X|+b|Y_i−Y|+c|Z_i−Z|
- a, b, c are the weighting coefficients of the three dimensions, which are used to control the weight of the three dimensions. In this embodiment, the weighting coefficients a, b, c are a=2, b=1 and c=3 respectively.

$\hat{A} = \frac{\sum_{i = 1}^{3} ω_{i} A_{i}}{\sum_{i = 1}^{3} ω_{i}}$

- 3) The attribute residual value R of the current point is decoded from the bitstream;
- 4) According to the attribute prediction value of the current point and the decoded attribute residual value, the reconstructed attribute value of the current point is A=Â+R.

Embodiment 23

A method for generating a new processing order in this embodiment includes the following contents.

For the reconstructed point cloud geometry, the Hilbert code corresponding to it is generated according to its three-dimensional spatial coordinates. As mentioned above, for a given three-dimensional coordinate (X₁, X₂, X₃), the coordinates of the point are expanded in binary as follows:

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

The Hilbert code HCode_N(X₁, X₂, X₃)=(C₁C₂. . . C_n. . . C_N−1C_N) of point (X₁, X₂, X₃) is obtained through the following process, The Hilbert code has 3*N binary bits. C_i, i=1, . . . , N, is the i-th substring code composed of three binary bits, while C₁is the highest 3-bit subcode and C_Nis the lowest 3-bit subcode. Hilbert codes are obtained by performing the following calculations:

$C_{1} = f_{S_{1}} (x_{1}^{1}, x_{2}^{1}, x_{3}^{1}), S_{1} \in [a, b, c, d, e, h, k, m, p, q, r, s]$

$C_{2} = f_{S_{2}} (x_{1}^{2}, x_{2}^{2}, x_{3}^{2}), S_{2} = g (S_{1}, C_{1})$

$\dots$

$C_{n} = f_{S_{n}} (x_{1}^{n}, x_{2}^{n}, x_{3}^{n}), S_{n} = g (S_{n - 1}, C_{n - 1})$

$\dots$

$C_{N} = f_{S_{N}} (x_{1}^{N}, x_{2}^{N}, x_{3}^{N}), S_{N} = g (S_{N - 1}, C_{N - 1})$

Among them, custom-character _iis the state when the subcode C_i−1of the i-th string is generated, and _iis one of the 12 states, as shown in FIG. 4. Each state corresponds to a conversion function (x₁ⁱ,x₂ⁱ,x₃ⁱ) of (x₁ⁱ,x₂ⁱ,x₃ⁱ) to C_i−1. ₁∈[a, b, c, d, e, h, k, m, p, q, r, s] means that when calculating C₁, one of the 12 states should be selected as the initial value state, and then each time custom-character _iis calculated according to the mapping relationship defined by the function g(_i−1, C_i−1), which contains the upper state _i−1corresponding to the subcode and the upper subcode C_i−1corresponding to the subcode. Until all subcodes C₁, C₂. . . C_Nare calculated and all subcodes are arranged in order, the Hilbert codeHCode_N(X₁, X₂, X₃) of points with three-dimensional space coordinates (x₁ⁱ,x₂ⁱ,x₃ⁱ) is obtained. The bits of the binary code of the Hilbert code cannot form a one-to-one correspondence with the bits of the binary code of the geometric coordinates.

The correspondence between the input x₁ⁱ,x₂ⁱ,x₃ⁱof a mapping function ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ) corresponding to the 12 states defined in FIG. 4 and the output C_iis as follows:

x₁ⁿ
x₂ⁿ
x₃ⁿ
C_i

0
0
0
000

0
0
1
001

0
1
0
011

0
1
1
010

1
0
0
111

1
0
1
110

1
1
0
100

The remaining functions are as follows:

ƒ_b(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_c(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_d(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_e(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_h(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_k(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_m(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_p(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_q(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_r(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_s(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

The transformation relationship between the above maps has the same meaning as the transformation relationship between the curves corresponding to the state, the order exchange of the binary coordinates input by ƒ(x₁ⁱ,x₂ⁱ,x₃ⁱ) corresponds to the symmetry of the plane in the curve, and the reverse of the binary coordinates input by ƒ(x₁ⁱ,x₂ⁱ,x₃ⁱ) corresponds to the rotation around the axis in the curve. The remaining mappings are shown in the following table:

C_i= f_b(x₁ⁱ, x₂ⁱ, x₃ⁱ)

C_i= f_c(x₁ⁱ, x₂ⁱ, x₃ⁱ)

x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i

0
0
0
000
0
0
0
000

0
0
1
011
0
0
1
111

0
1
0
111
0
1
0
001

0
1
1
100
0
1
1
110

1
0
0
001
1
0
0
011

1
0
1
010
1
0
1
100

1
1
0
110
1
1
0
010

1
1
1
101
1
1
1
101

C_i= f_d(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_e(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_h(x₁ⁱ, x₂ⁱ, x₃ⁱ)

x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i

0
0
0
010
0
0
0
100
0
0
0
110

0
0
1
011
0
0
1
111
0
0
1
001

0
1
0
001
0
1
0
011
0
1
0
111

0
1
1
000
0
1
1
000
0
1
1
000

1
0
0
101
1
0
0
101
1
0
0
101

1
0
1
100
1
0
1
110
1
0
1
010

1
1
0
110
1
1
0
010
1
1
0
100

1
1
1
111
1
1
1
001
1
1
1
011

C_i= f_k(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_m(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_p(x₁ⁱ, x₂ⁱ, x₃ⁱ)

x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i

0
0
0
110
0
0
0
010
0
0
0
100

0
0
1
111
0
0
1
001
0
0
1
011

0
1
0
101
0
1
0
101
0
1
0
101

0
1
1
100
0
1
1
110
0
1
1
010

1
0
0
001
1
0
0
011
1
0
0
111

1
0
1
000
1
0
1
000
1
0
1
000

1
1
0
010
1
1
0
100
1
1
0
110

1
1
1
011
1
1
1
111
1
1
1
001

C_i= f_q(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_r(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_s(x₁ⁱ, x₂ⁱ, x₃ⁱ)

x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i

0
0
0
100
0
0
0
110
0
0
0
010

0
0
1
101
0
0
1
101
0
0
1
101

0
1
0
111
0
1
0
001
0
1
0
011

0
1
1
110
0
1
1
010
0
1
1
100

1
0
0
011
1
0
0
111
1
0
0
001

1
0
1
010
1
0
1
100
1
0
1
110

1
1
0
000
1
1
0
000
1
1
0
000

1
1
1
001
1
1
1
011
1
1
1
111

The function custom-character _i=g(_i−1, C_i−1) is used to obtain the state _iof the current level from the previous level state _i−1and the previous level subcode C_i−1, which is shown as follows:

- Table 1. the g function mapping relation to get _iaccording to _i−1and C_i−1

_i−1

C_i−1
a
b
c
d
e
h
k
m
p
q
r
s

000
c
a
b
h
d
e
p
k
m
s
q
r

001
b
c
a
e
h
d
m
p
k
r
s
q

010
b
c
a
e
h
d
m
p
k
r
s
q

011
q
e
p
k
b
s
d
r
c
a
m
h

100
q
e
p
k
b
s
d
r
c
a
m
h

101
m
s
d
r
p
a
b
h
q
e
c
k

110
m
s
d
r
p
a
b
h
q
e
c
k

111
h
k
r
c
q
m
s
a
e
p
d
b

In this example, the numbers corresponding to the 12 states shown in FIG. 4 can be changed, but the g( custom-character _i−1, C_i−1) function in Table 1 should also be adjusted accordingly so that the state transition relationship of _iobtained from _i−1) remains unchanged.

Embodiment 24

A method of generating a new processing order for this embodiment includes the following

For the reconstructed point cloud geometry, the corresponding Hilbert code is generated according to its three-dimensional space coordinates. As mentioned above, for a given three-dimensional coordinate of (X₁, X₂, X₃), its coordinates are expanded in binary as follows

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

Among them, custom-character _iis the state when the subcode C_i−1of the i-th string is generated, and _iis one of the 12 states, as shown in FIG. 4. Each state corresponds to a conversion function (x₁ⁱ,x₂ⁱ,x₃ⁱ) of (x₁ⁱ,x₂ⁱ,x₃ⁱ) to C_i−1. _i∈[a, b, c, d, e, h, k, m, p, q, r, s] means that when calculating C₁, one of the 12 states should be selected as the initial value state, and then each time custom-character _iis calculated according to the mapping relationship defined by the function g(_i−1, C_i−1), which contains the upper state _i−1corresponding to the subcode and the upper subcode C_i−1corresponding to the subcode. Until all subcodes C₁, C₂. . . C_Nare calculated and all subcodes are arranged in order, the Hilbert codeHCode_N(X₁, X₂, X₃) of points with three-dimensional space coordinates (x₁ⁱ,x₂ⁱ,x₃ⁱ) is obtained. The bits of the binary code of the Hilbert code cannot form a one-to-one correspondence with the bits of the binary code of the geometric coordinates.

The correspondence between the input x₁ⁱ,x₂ⁱ,x₃ⁱof a mapping function ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ) corresponding to the 12 states defined in FIG. 4 and the output C_iis as follows:

c
₁
ⁱ
=x
₁
ⁱ

c
₂
ⁱ
=x
₁
ⁱ
−x
₂
ⁱ

c
₃
ⁱ=x₁^l(x₂ⁱ⊕x₃ⁱ)+x₁ⁱ(x₂ⁱ⊙x₃ⁱ)

C
_i
=c
₁
ⁱ
c
₂
ⁱ
c
₃
ⁱ

Among them, ⊕ denotes the exclusive-or operation, and ⊙ denotes the inclusive-or operation. The mapping relationship between the remaining states and ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ) has the following operational relationship, where x denotes the reverse operation of x.

ƒ_b(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_c(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_d(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_e(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_h(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_k(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_m(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_p(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_q(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_r(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_s(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

The mapping relationship of the function custom-character _i=g(_i−1, C_i−1) is shown in Table 2.

TABLE 2

the g function mapping relation to

get custom-character

_iaccording to custom-character

_i−1and C_i−1

custom-character

In this embodiment, the numbers corresponding to the 12 states shown in FIG. 4 can be changed, but the g( custom-character _i−1, C_i−1) function in Table 2 should also be adjusted accordingly so that the state transition relationship of _iobtained from _i−1) remains unchanged.

Embodiment 25

A method of generating a new processing order for this embodiment includes the following

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

The correspondence between the input x₁ⁱ,x₂ⁱ,x₃ⁱand the output C_iof a mapping function ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ) is as follows:

C_i= f_a(x₁ⁱ, x₂ⁱ, x₃ⁱ)

x₁ⁿ
x₂ⁿ
x₃ⁿ
C_i

0
0
0
101

0
0
1
100

0
1
0
110

0
1
1
111

1
0
0
010

1
0
1
011

1
1
0
001

1
1
1
000

The remaining functions are as follows:

ƒ_b(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_c(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_d(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_e(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_h(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_k(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_m(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_p(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_q(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_r(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_s(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

The remaining mappings are shown in the following table:

C_i= f_b(x₁ⁱ, x₂ⁱ, x₃ⁱ)

C_i= f_c(x₁ⁱ, x₂ⁱ, x₃ⁱ)

x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i

0
0
0
101
0
0
0
101

0
0
1
110
0
0
1
010

0
1
0
010
0
1
0
100

0
1
1
001
0
1
1
011

1
0
0
100
1
0
0
110

1
0
1
111
1
0
1
001

1
1
0
011
1
1
0
111

1
1
1
000
1
1
1
000

C_i= f_d(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_e(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_h(x₁ⁱ, x₂ⁱ, x₃ⁱ)

x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i

0
0
0
111
0
0
0
001
0
0
0
011

0
0
1
110
0
0
1
010
0
0
1
100

0
1
0
100
0
1
0
110
0
1
0
010

0
1
1
101
0
1
1
101
0
1
1
101

1
0
0
000
1
0
0
000
1
0
0
000

1
0
1
001
1
0
1
011
1
0
1
111

1
1
0
011
1
1
0
111
1
1
0
001

1
1
1
010
1
1
1
100
1
1
1
110

C_i= f_k(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_m(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_p(x₁ⁱ, x₂ⁱ, x₃ⁱ)

x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i

0
0
0
011
0
0
0
111
0
0
0
001

0
0
1
010
0
0
1
100
0
0
1
110

0
1
0
000
0
1
0
000
0
1
0
000

0
1
1
001
0
1
1
011
0
1
1
111

1
0
0
100
1
0
0
110
1
0
0
010

1
0
1
101
1
0
1
101
1
0
1
101

1
1
0
111
1
1
0
001
1
1
0
011

1
1
1
110
1
1
1
010
1
1
1
100

C_i= f_q(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_r(x₁ⁱ, x₂ⁱ, x₃ⁱ)
C_i= f_s(x₁ⁱ, x₂ⁱ, x₃ⁱ)

x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i
x₁ⁱ
x₂ⁱ
x₃ⁱ
C_i

0
0
0
001
0
0
0
011
0
0
0
111

0
0
1
000
0
0
1
000
0
0
1
000

0
1
0
010
0
1
0
100
0
1
0
110

0
1
1
011
0
1
1
111
0
1
1
001

1
0
0
110
1
0
0
010
1
0
0
100

1
0
1
111
1
0
1
001
1
0
1
011

1
1
0
101
1
1
0
101
1
1
0
101

1
1
1
100
1
1
1
110
1
1
1
010

TABLE 3

the g function mapping relation to

get custom-character

_iaccording to custom-character

_i−1and C_i−1

custom-character

In this embodiment, the numbers corresponding to the 12 states can be changed, but the g( custom-character _i−1, C_i−1) function in Table 3 should also be adjusted accordingly so that the state transition relationship of _iobtained from S_i−1) remains unchanged.

Embodiment 26

A method of generating a new processing order for this embodiment includes the following

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

Among them, custom-character _iis the state when the subcode C_i−1of the i-th string is generated, and _iis one of the 12 states, as shown in FIG. 4. Each state corresponds to a conversion function (x₁ⁱ,x₂ⁱ,x₃ⁱ) of (x₁ⁱ,x₂ⁱ,x₃ⁱ) to C_i−1. _i∈[a, b, c, d, e, h, k, m, p, q, r, s] means that when calculating C₁, one of the 12 states should be selected as the initial value state, and then each time custom-character _iis calculated according to the mapping relationship defined by the function g(_i−1, C_i−1), which contains the upper state _i−1corresponding to the subcode and the upper subcode C_i−1corresponding to the subcode. Until all subcodes C₁, C₂. . . C_Nare calculated and all subcodes are arranged in order, the Hilbert codeHCode_N(X₁, X₂, X₃) of points with three-dimensional space coordinates (x₁ⁱ,x₂ⁱ,x₃ⁱ) is obtained. The bits of the binary code of the Hilbert code cannot form a one-to-one correspondence with the bits of the binary code of the geometric coordinates.

The correspondence between the input x₁ⁱ,x₂ⁱ,x₃ⁱof a mapping function ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ) corresponding to the 12 states defined in FIG. 4 and the output C_iis as follows:

c
₁
ⁱ
=x
₁
ⁱ

c
₂
ⁱ
=x
₁
ⁱ
−x
₂
ⁱ

c
₃
ⁱ=x₁^l(x₂ⁱ⊕x₃ⁱ)+x₁ⁱ(x₂ⁱ⊙x₃ⁱ)

C
_i
=c
₁
ⁱ
c
₂
ⁱ
c
₃
ⁱ

ƒ_b(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_c(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_d(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_e(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_h(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_k(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_m(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_p(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_q(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_r(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_s(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

The transformation relationship between the above maps has the same meaning as the transformation relationship between the curves corresponding to the state The order exchange of the binary coordinates input by ƒ(x₁ⁱ,x₂ⁱ,x₃ⁱ) corresponds to the symmetry of the plane in the curve, and the reverse of the binary coordinates input by ƒ(x₁ⁱ,x₂ⁱ,x₃ⁱ) corresponds to the rotation around the axis in the curve.

The mapping relationship of the function custom-character _i=g(_i−1, C_i−1) is shown in Table 4.

TABLE 4

the g function mapping relation to

get custom-character

_iaccording to custom-character

_i−1and C_i−1

custom-character

In this example, the numbers corresponding to the 12 states can be changed, but the g ( custom-character _i−1, C_i−1) function in Table 4 should also be adjusted accordingly so that the state transition relationship of _iobtained from _i−1) remains unchanged.

Embodiment 27

A point cloud attribute decoding method based on a new processing order for this embodiment includes the following contents.

According to the three-dimensional geometric coordinates of all points in the point set, the corresponding Hilbert code is calculated by any method in embodiment 23 to embodiment 26, and then these points are reordered according to the Hilbert code values of these points, and the multiple points arranged in Hilbert order are obtained. The bits of the binary code of the Hilbert code cannot form a one-to-one correspondence with the bits of the binary code of the geometric coordinates.

The attribute residual values of multiple points are obtained by decoding from the bitstream.

The attribute residual values of these points correspond to the Hilbert codes of each point in the Hilbert order point set, and also correspond to the three-dimensional geometric coordinates of each point in the Hilbert order. That is, the attribute residuals r₁, r₂, . . . decoded from the stream correspond to the points P₁, P₂, . . . sorted after Hilbert order.

Embodiment 28

A point cloud attribute decoding method based on a new processing order for this embodiment includes the following contents.

In this paper, any method of embodiments 1 to 22 is used to decode the attribute residual value from the bitstream and predict the attribute value of the decoding point, and then the reconstructed attribute value of the point is obtained according to the attribute residual value corresponding to the point.

Embodiment 29

A point cloud attribute encoding method based on a new processing order in this embodiment includes the following contents.

Use any attribute prediction method of implementation examples 1 to 22 to predict the attribute value of the coding point, and then obtain the attribute residual value of the point according to the actual attribute value corresponding to the point.

Write the attribute residual values of multiple points to the bitstream.

Embodiment 30

A device for generating a new processing order of this embodiment includes the following contents.

Subcode generation module: For the reconstructed point cloud geometry, for a given three-dimensional coordinate point (X₁, X₂, X₃), its coordinates are expanded in binary as follows:

X
₁=(x₁¹x₁². . . x₁ⁿ. . . x₁^N−1x₁^N),X₂=(x₂¹x₂². . . x₂ⁿ. . . x₂^N−1x₂^N),X₃=(x₃¹x₃². . . x₃ⁿ. . . x₃^N−1x₃^N)

Input the binary bits x₁ⁱ,x₂ⁱ,x₃ⁱof its three-dimensional space coordinates, and output the corresponding Hilbert subcode C_i, i=1, . . . , N, where the subcode is composed of three binary bits, while C₁is the highest 3-bit subcode and C_Nis the lowest 3-bit subcode. Hilbert codes are obtained by performing the following calculations:

C
_i= custom-character (x₁ⁱ,x₂ⁱ,x₃ⁱ)

State transition module:

$S_{i} = {\begin{matrix} take any of the values in [a, b, c, d, e, h, k, m, p, q, r, s], & i = 1 \\ g (S_{i - 1}, C_{i - 1}), & i \neq 1 \end{matrix}$

Hilbert code generation module: until all subcodes C₁, C₂. . . C_Nare calculated and all subcodes are arranged in order, the Hilbert code HCode_N(X₁, X₂, X₃) of points with three-dimensional space coordinates (x₁ⁱ,x₂ⁱ,x₃ⁱ) is obtained. The bits of the binary code of the Hilbert code cannot form a one-to-one correspondence with the bits of the binary code of the geometric coordinates.

The correspondence between the input x₁ⁱ,x₂ⁱ,x₃ⁱand the output C_iof a mapping function ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ) defined in FIG. 4 is as follows:

C_i= f_a(x₁ⁱ, x₂ⁱ, x₃ⁱ)

x₁ⁿ
x₂ⁿ
x₃ⁿ
C_i

0
0
0
000

0
0
1
001

0
1
0
011

0
1
1
010

1
0
0
111

1
0
1
110

1
1
0
100

1
1
1
101

The remaining functions are as follows:

ƒ_b(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_c(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_d(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_e(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_h(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_k(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_m(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_p(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

ƒ_q(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₁ⁱ,x₂ⁱ,x₃ⁱ),ƒ_r(x₁ⁱ,x₂ⁱ,x₃ⁱ)=ƒ_a(x₂ⁱ,x₃ⁱ,x₁ⁱ),ƒ_s(x₁ⁱ,x₂ⁱ,x₃ⁱ)=

ƒ_a(x₃ⁱ,x₁ⁱ,x₂ⁱ),

The remaining mappings are shown in the following table:

TABLE 5

the g function mapping relation to

get custom-character

_iaccording to custom-character

_i−1and C_i−1

custom-character

In this embodiment, the numbers corresponding to the 12 states shown in FIG. 4 can be changed, but the g( custom-character _i−1, C_i−1) function in Table 5 should also be adjusted accordingly so that the state transition relationship of _iobtained from _i−1) remains unchanged.

Embodiment 31

A device for generating a new processing order of this embodiment includes the following contents.

Using the subcode generation module, state transition module and Hilbert code generation module of embodiment 30, the corresponding Hilbert code is output for the input three-dimensional coordinates of (X₁, X₂, X₃).

The Hilbert sorting module: input the Hilbert code of multiple points, reorder these points according to the Hilbert code value of these points, and output the multiple points arranged in Hilbert order. The bits of the binary code of the Hilbert code cannot form a one-to-one correspondence with the bits of the binary code of the geometric coordinates;

according to the attribute residual decoding module of embodiments 6 to 10 and 20 to 22, the attribute residual values of the multiple points are obtained by decoding from the stream.

The attribute residual corresponding module: the respective attribute residual values of the multiple points correspond to the Hilbert code of each point in the Hilbert order, and also correspond to the three-dimensional geometric coordinates of each point in the Hilbert order.

According to the attribute prediction value acquisition module and the reconstructed attribute value acquisition module of implementation examples 6 to 10 and implementation examples 20 to 22, the input of the attribute prediction value acquisition module is the reconstructed point cloud, and the attribute prediction value of these points is output. The reconstructed attribute value acquisition module inputs the attribute prediction value and the attribute residual value corresponding to the reconstructed point to obtain the reconstructed attribute value of the current point.

Finally, we need to explain all the above implementation methods, such as the implementation methods of custom-character (x₁ⁱ,x₂ⁱ,x₃ⁱ), g(_i−1, C_i−1), including but not limited to functions, lookup tables, fixed circuit implementations, or state transitions using state machines.

Number	Date	Country	Kind
202010822352.0	Aug 2020	CN	national
202011063672.9	Sep 2020	CN	national
202011452464.8	Dec 2020	CN	national

POINT CLOUD ATTRIBUTE ENCODING AND DECODING METHOD AND DEVICE

Information

Publication Number

Date Filed

Date Published

Inventors

Original Assignees

CPC

International Classifications

Abstract

Description

Claims

Priority Claims (3)

PCT Information