Classifying and controlling encoding quality for textured, dark smooth and smooth video content

Description

BACKGROUND

Engineers use compression (also called coding or encoding) to reduce the bit rate of digital video. Compression decreases the cost of storing and transmitting video by converting the video into a lower bit rate form. Decompression (also called decoding) reconstructs a version of the original video from the compressed form. A “codec” is an encoder/decoder system.

Intra-picture Compression and Inter-picture Compression

In general, video compression techniques include “intra-picture” compression and “inter-picture” compression. Intra-picture compression techniques compress an individual picture without reference to other pictures that have been compressed and reconstructed. Inter-picture compression techniques compress a picture with reference to preceding and/or following picture(s) (often called reference or anchor pictures) that have already been compressed and reconstructed.

Most encoders use a frequency transform during intra-picture compression and inter-picture compression. For example, the encoder splits a picture into non-overlapping blocks of samples and applies a forward frequency transform to individual blocks. The frequency transform maps the sample values of a block to transform coefficients, which are coefficients of basis functions that correspond to frequency components. In particular, the lowest frequency coefficient—called the DC coefficient—indicates the average sample value for the block. The other coefficients—called AC coefficients—indicate patterns of changes in sample values of the block, from gradual low-frequency variations across the block to sharper high-frequency variations within the block. In many encoding scenarios, a relatively small number of frequency coefficients (e.g., the DC coefficient and lower frequency AC coefficients) capture much of the energy or signal content in the block. The encoder quantizes the transform coefficients, resulting in a block of quantized transform coefficients. The encoder further encodes the quantized transform coefficients, for example, using entropy coding, and outputs a bitstream of compressed video information.

In corresponding decoding, a decoder reads the bitstream of compressed video information and performs operations to reconstruct the pictures that were encoded. When the encoding uses lossy compression (e.g., in quantization), the reconstructed pictures approximate the source pictures that were encoded but are not exactly the same. For example, to reconstruct a version of the original 8×8 block of an intra-compressed picture, the decoder reconstructs quantized transform coefficients using entropy decoding. The decoder inverse quantizes the quantized transform coefficients of the block and applies an inverse frequency transform to the de-quantized transform coefficients, producing the reconstructed version of the original 8×8 block.

Inter-picture compression techniques often use motion compensation to reduce bit rate by exploiting temporal redundancy in video. In general, motion compensation is a process of producing predictions from reference picture(s) (such as previously encoded/decoded picture(s)) using motion data. An encoder and decoder store previously coded/decoded pictures in a picture store. The reference pictures in the picture store can then provide motion-compensated predictor blocks for the blocks of a current picture being encoded. Often, the encoder does not find a perfect match. For this reason, the encoder computes the sample-by-sample differences between the current block and its motion-compensated prediction to determine a residual (also called error signal). The residual is frequency transformed, quantized, and entropy encoded. When reconstructing residuals, a decoder (and also the encoder) reconstructs transform coefficients that were quantized and performs an inverse frequency transform. The decoder/encoder performs motion compensation to compute motion-compensated predictors, and combines the predictors with the residuals.

Types of Quantization

According to one possible definition, quantization is a term used for an approximating non-reversible mapping function commonly used for lossy compression, in which there is a specified set of possible output values, and each member of the set of possible output values has an associated set of input values that result in the selection of that particular output value. A variety of quantization techniques have been developed, including scalar or vector, uniform or non-uniform, and adaptive or non-adaptive quantization.

According to one possible definition, a scalar quantizer is an approximating functional mapping x→Q[x] of an input value x to a quantized value Q[x], sometimes called a reconstructed value. FIG. 1 shows a “staircase” I/O function (100) for a scalar quantizer, along with example reconstruction points for inverse quantization. The horizontal axis is a number line for an input variable x, and the vertical axis indicates the corresponding quantized values Q[x]. The number line is partitioned by thresholds such as the threshold (110). Each value of x within a given range between a pair of adjacent thresholds is assigned the same quantized value Q[x]. For example, each value of x within the range (120) is assigned the same quantized value (130). (At a threshold, one of the two possible quantized values is assigned to an input x, depending on the system.) Overall, the quantized values Q[x] exhibit a discontinuous, staircase pattern. The placement of the thresholds on the number line may be uniformly spaced (as shown in FIG. 1) or non-uniformly spaced.

A scalar quantizer can be decomposed into two distinct stages. The first stage is the classifier stage, in which a classifier function mapping x→A [x] maps an input x to a quantization index A[x], which is often integer-valued. In essence, the classifier segments an input number line or data set, as in FIG. 1, by thresholds such as the threshold (110).

In the second stage, a reconstructor functional mapping k→β[k] maps each quantization index k to a reconstruction value β[k]. In essence, the reconstructor selects a value for reconstruction of each region determined by the classifier. The reconstructor functional mapping may be implemented, for example, using a lookup table. FIG. 1 shows (as open circles) example reconstruction points according to a midpoint reconstruction rule. Overall, the classifier relates to the reconstructor as follows:

Q[x]=β[A[x]] (1).

In common usage, the term “quantization” is often used to describe the classifier stage, which is performed during encoding. The term “inverse quantization” is similarly used to describe the reconstructor stage, whether performed during encoding or decoding.

A non-uniform quantizer has threshold values that are not uniformly spaced for all classifier regions. According to one possible definition, a dead zone plus uniform threshold quantizer (“DZ+UTQ”) is a quantizer with uniformly spaced threshold values for all classifier regions except the one containing the zero input value (which is called the dead zone (“DZ”)). In a general sense, a DZ+UTQ is a non-uniform quantizer, since the DZ size is different than the other classifier regions.

FIG. 2 shows a staircase I/O function (200) for a DZ+UTQ, in which the DZ is wider than the other steps s. The number line is partitioned by thresholds such as the threshold (210), and each value of x within a given range between a pair of adjacent thresholds is assigned the same quantized value Q[x]. For example, each value of x within the range (220) is assigned the same quantized value (230). In FIG. 2, the DZ is twice as wide as the other classification zones. FIG. 2 shows (as open circles) example reconstruction points according to a midpoint reconstruction rule.

Adjusting Quantization

Quantization can affect the fidelity with which transform coefficients are encoded, which in turn can affect bit rate. Coarser quantization tends to decrease fidelity to the original transform coefficients (and produce more distortion) as the coefficients are more coarsely approximated. Bit rate also decreases, however, when decreased complexity can be exploited with lossless compression (e.g., entropy encoding). Conversely, finer quantization tends to preserve fidelity and quality (and produce less distortion) but result in higher bit rates.

Some encoders adjust quantization between pictures and/or within pictures to control where distortion is introduced. For a given bit rate/quality level, this allows an encoder to introduce more distortion where it will be less visible and/or avoid introducing distortion where it would be more visible. The allocation of available bits among pictures and within pictures plays an important role in how distortion is introduced and how the user perceives the quality of the video.

Different encoders typically apply different quantization rules, but there are some common principles. Quantization can produce visible artifacts that tend to be more artificial-looking and visually distracting than simple loss of fine detail. For example, the human visual system is more sensitive to distortion in relatively smooth content than to distortion in textured content. High texture levels tend to mask quality degradation and quantization artifacts. On the other hand, in regions with lower texture levels, distortion tends to be more visible. So, in smooth regions distortion may create a visible line, step or other flaw in the reconstructed image, while the same amount of distortion may not create noticeable flaws in textured areas due to masking effects of surrounding detail.

Thus, a common strategy is to allocate relatively more bits to smooth content and relatively fewer bits to textured content, so that less distortion is introduced in smooth content at the expense of more distortion in the textured content (where the distortion is not as perceptually noticeable). To identify textured content and non-textured content, various texture metrics and texture thresholds have been used. In some cases, an encoder varies quantization depending on texture. This allows the encoder to coarsen quantization when doing so will not dramatically increase perceptibility of the distortion and use finer quantization in other situations.

Standards and product specifications that focus only on achieving interoperability will often specify reconstruction values for inverse quantization without specifying a classification rule for quantization. In other words, some specifications may define the functional mapping k→β[k] for reconstruction without defining the functional mapping x→A[x] for classification. This allows a decoder built to comply with the standard/product to reconstruct information correctly. In contrast, encoders are often given the freedom to change the classifier. For classification, the thresholds can be defined so that certain input values will be mapped to more common (and hence, lower bit rate) indices, which makes the reconstruction values closer to optimal for some content. This also allows the encoder to adjust to expected distributions in values. For example, an encoder may define the DZ threshold to be wider or narrower for a quantizer. Or, more generally, the encoder may define other thresholds according to which values are quantized so as to quantize values more aggressively.

The preceding adaptive quantization mechanisms help improve performance in many scenarios. In some scenarios, however, they fail to provide quantization control that is both usable and sufficiently fine-grained. For example, in some scenarios, previous adaptive quantization mechanisms provide insufficient control over how content is classified as textured or non-textured. As a result, encoding of smooth areas introduces an unacceptable amount of distortion. Another problem is that, in some scenarios, previous adaptive quantization mechanisms provide insufficient control over bit allocation for different types of non-textured content. Given the critical importance of video compression to digital video, it is not surprising that video compression is a richly developed field. Whatever the benefits of previous video compression techniques, however, they do not have the advantages of the following techniques and tools.

SUMMARY

In summary, the present application is directed to techniques and tools for content classification and adaptive quantization. In some embodiments, a video encoder classifies blocks of video pictures as textured, dark smooth or smooth and adjusts quantization of the blocks depending on the classifications. This helps the encoder control how distortion is introduced in textured regions, dark smooth regions, and other smooth regions, improving overall perceived quality.

According to a first aspect of the techniques and tools, a tool such as an encoder classifies units such as blocks of a picture. For each of multiple units of a picture, the tool assigns a classification to the unit based at least in part on the results of analyzing the content of the unit. The tool then uses the classifications to process the units. The classification for at least one of the units is a first classification type, and the classification for at least one of the plural units is a second classification type indicating darker content than the first classification type. In an example implementation, a video encoder classifies a block as textured or non-textured by comparing the energy of AC coefficients for the block to a texture threshold. If the block is not textured, the video encoder further classifies the block as dark smooth or smooth depending on average intensity value for the block. Using the block classifications and one or more control parameters to control bit allocation for dark smooth content relative to other smooth content, the encoder encodes the blocks and outputs encoded video information.

According to a second aspect of the techniques and tools, an encoding tool encodes units such as blocks of video pictures. The encoding tool sets one or more parameters to control bit allocation for a first classification type relative to a second classification type during encoding, where the second classification type indicates darker content than the first classification type. The parameter(s) can include one or more target AC coefficient counts for the first classification type and one or more target AC coefficient counts for the second classification type. To set the parameter(s), the tool encodes (in a first encoding pass) the units using initial values for the parameter(s) then adjusts the values of the parameter(s) based at least in part upon results of the first pass encoding with the initial values. For example, the tool changes the initial values depending upon target bit rate and resulting bit rate of the first pass encoding. The tool then encodes (in a second encoding pass) the units using the adjusted values and outputs the encoded units in a bit stream. The first pass encoding and second pass encoding can include classification of the blocks as textured, dark smooth and other smooth, and can use an adaptive, user-adjustable texture threshold.

According to third aspect of the techniques and tools, an encoding tool includes a user interface module, an encoding control module, and one or more video encoding modules. The user interface module is adapted to receive user input that indicates a level of texture classification. The encoding control module is adapted to set a texture threshold based at least in part upon the level of texture classification, using a non-linear mapping of possible values of level of texture classification to the possible values of the texture threshold. The non-linear mapping can embody finer gradations for simpler content (such as animation content) and coarser gradations for more complex content (such as natural video), so as to improve control over textured/non-textured threshold decisions. The video encoding module(s) can include a module adapted to set parameters to control bit allocation between textured content, dark smooth content, and other smooth content during the encoding.

The foregoing and other objects, features, and advantages of the invention will become more apparent from the following detailed description, which proceeds with reference to the accompanying figures.

BRIEF DESCRIPTION OF THE DRAWINGS

FIGS. 1 and 2 are charts illustrating aspects of quantization according to the prior art.

FIG. 3 is a block diagram of a suitable computing environment in which several described embodiments may be implemented.

FIG. 4 is a block diagram of a video encoder system in conjunction with which several described embodiments may be implemented.

FIG. 5 is a flowchart illustrating a generalized technique for using texture thresholds separated by non-linear gradations when classifying content.

FIG. 6 is a flowchart illustrating a generalized technique for classifying non-textured content as smooth or dark smooth.

FIG. 7 is a flowchart illustrating an example technique for classifying blocks as textured, smooth or dark smooth.

FIG. 8 is a flowchart illustrating a generalized technique for allocating bits between textured video content, dark smooth video content and other smooth video content.

FIGS. 9 and 10 are charts illustrating aspects of quantization in some embodiments.

FIG. 11 is a chart illustrating an example block of transform coefficients, and FIGS. 12A and 12B are charts illustrating example quantization decisions for the transform coefficients shown in FIG. 11.

FIG. 13 is a flowchart illustrating an example technique for setting quantization parameter values and DZ size for quantization of macroblocks with dark smooth or other smooth blocks.

FIG. 14 is a flowchart illustrating a generalized technique for setting relative bit allocations for textured video content, dark smooth video content and other smooth video content.

DETAILED DESCRIPTION

The present application relates to techniques and tools for improving quantization by controlling encoding quality of dark smooth video content relative to other smooth video content and textured video content. Certain types of quantization artifacts are particularly noticeable in dark areas that lack texture detail. By allocating relatively more bits to dark smooth areas, overall perceived quality can be improved.

In some embodiments, an encoding tool classifies blocks of video as smooth, dark smooth or textured. The encoding tool allocates bits to blocks in the three different categories. For example, the encoding tool changes macroblock quantization step sizes and/or quantizer DZ size depending on block classifications. Then, when the encoding tool introduces distortion to regulate bit rate, it tends to introduce more distortion where it is expected to be less perceptible (e.g., textured areas) and introduce less distortion where it would likely be more perceptible (e.g., dark smooth areas), thereby helping to achieve a desired perceptual quality within a given bit rate constraint.

In some embodiments, the encoding tool uses a rate control mechanism to help achieve the desired allocations during encoding. The encoding tool can use a two-pass encoding process to help achieve the desired perceptual quality of compressed video via allocation of bits between smooth, dark smooth, and textured content. In the first pass, the encoding tool encodes the video using initial settings for control parameters. The encoding tool then analyzes results of the first encoding pass, so as to determine the desired settings for the control parameters for encoding in a second pass. In the second pass, the encoding tool encodes the video using the desired settings for the control parameters.

Various alternatives to the implementations described herein are possible. Certain techniques described with reference to flowchart diagrams can be altered by changing the ordering of stages shown in the flowcharts, by repeating or omitting certain stages, etc. For example, initial stages of classification (e.g., obtaining frequency coefficients for a picture, macroblock(s) or block(s)) can be completed before later stages begin, or operations for the different stages can be interleaved on a block-by-block, macroblock-by-macroblock, or other region-by-region basis.

The various techniques and tools described herein can be used in combination or independently. Different embodiments implement one or more of the described techniques and tools. Aside from uses in video compression, the adaptive quantization techniques and tools can be used in image compression, other compression, or other areas. While many examples described herein involve quantization of AC coefficients for blocks by a video encoder, alternatively the techniques and tools described herein are applied to quantization of AC coefficients for blocks in an image encoder.

Some of the techniques and tools described herein address one or more of the problems noted in the background. Typically, a given technique/tool does not solve all such problems. Rather, in view of constraints and tradeoffs in encoding time, resources, and/or quality, the given technique/tool improves encoding performance for a particular implementation or scenario.

I. Computing Environment.

FIG. 3 illustrates a generalized example of a suitable computing environment (300) in which several of the described embodiments may be implemented. The computing environment (300) is not intended to suggest any limitation as to scope of use or functionality, as the techniques and tools may be implemented in diverse general-purpose or special-purpose computing environments.

With reference to FIG. 3, the computing environment (300) includes at least one processing unit (310) and memory (320). In FIG. 3, this most basic configuration (330) is included within a dashed line. The processing unit (310) executes computer-executable instructions and may be a real or a virtual processor. In a multi-processing system, multiple processing units execute computer-executable instructions to increase processing power. The memory (320) may be volatile memory (e.g., registers, cache, RAM), non-volatile memory (e.g., ROM, EEPROM, flash memory, etc.), or some combination of the two. The memory (320) stores software (380) implementing an encoding tool that implements one or more of the described techniques for controlling encoding quality for textured, dark smooth and other smooth video content.

A computing environment may have additional features. For example, the computing environment (300) includes storage (340), one or more input devices (350), one or more output devices (360), and one or more communication connections (370). An interconnection mechanism (not shown) such as a bus, controller, or network interconnects the components of the computing environment (300). Typically, operating system software (not shown) provides an operating environment for other software executing in the computing environment (300), and coordinates activities of the components of the computing environment (300).

The storage (340) may be removable or non-removable, and includes magnetic disks, magnetic tapes or cassettes, CD-ROMs, DVDs, or any other medium which can be used to store information and which can be accessed within the computing environment (300). The storage (340) stores instructions for the software (380) implementing the encoding tool.

The input device(s) (350) may be a touch input device such as a keyboard, mouse, pen, or trackball, a voice input device, a scanning device, or another device that provides input to the computing environment (300). For audio or video encoding, the input device(s) (350) may be a sound card, video card, TV tuner card, or similar device that accepts audio or video input in analog or digital form, or a CD-ROM or CD-RW that reads audio or video samples into the computing environment (300). The output device(s) (360) may be a display, printer, speaker, CD-writer, or another device that provides output from the computing environment (300).

The communication connection(s) (370) enable communication over a communication medium to another computing entity. The communication medium conveys information such as computer-executable instructions, audio or video input or output, or other data in a modulated data signal. A modulated data signal is a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media include wired or wireless techniques implemented with an electrical, optical, RF, infrared, acoustic, or other carrier.

The techniques and tools can be described in the general context of computer-readable media. Computer-readable media are any available media that can be accessed within a computing environment. By way of example, and not limitation, with the computing environment (300), computer-readable media include memory (320), storage (340), communication media, and combinations of any of the above.

The techniques and tools can be described in the general context of computer-executable instructions, such as those included in program modules, being executed in a computing environment on a target real or virtual processor. Generally, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The functionality of the program modules may be combined or split between program modules as desired in various embodiments. Computer-executable instructions for program modules may be executed within a local or distributed computing environment.

For the sake of presentation, the detailed description uses terms like “classify” and “set” to describe computer operations in a computing environment. These terms are high-level abstractions for operations performed by a computer, and should not be confused with acts performed by a human being. The actual computer operations corresponding to these terms vary depending on implementation.

II. Generalized Video Encoder.

FIG. 4 is a block diagram of a generalized video encoder (400) in conjunction with which some described embodiments may be implemented. The encoder (400) receives a sequence of video pictures including a current picture (405) and produces compressed video information (495) as output to storage, a buffer, or a communications connection. The format of the output bitstream can be a Windows Media Video or SMPTE 421M format, MPEG-x format (e.g., MPEG-1, MPEG-2, or MPEG-4), H.26× format (e.g., H.261, H.262, H.263, or H.264), or other format.

The encoder (400) processes video pictures. The term picture generally refers to source, coded or reconstructed image data. For progressive video, a picture is a progressive video frame. For interlaced video, a picture may refer to an interlaced video frame, the top field of the frame, or the bottom field of the frame, depending on the context. The encoder (400) is block-based and uses a 4:2:0 macroblock format for frames, with each macroblock including four 8×8 luma blocks (at times treated as one 16×16 macroblock) and two 8×8 chroma blocks. For fields, the same or a different macroblock organization and format may be used. The 8×8 blocks may be further sub-divided at different stages, e.g., at the frequency transform and entropy encoding stages. The encoder (400) can perform operations on sets of samples of different size or configuration than 8×8 blocks and 16×16 macroblocks. Alternatively, the encoder (400) is object-based or uses a different macroblock or block format.

Returning to FIG. 4, the encoder system (400) compresses predicted pictures and intra-coded, key pictures. For the sake of presentation, FIG. 4 shows a path for key pictures through the encoder system (400) and a path for predicted pictures. Many of the components of the encoder system (400) are used for compressing both key pictures and predicted pictures. The exact operations performed by those components can vary depending on the type of information being compressed.

A predicted picture (e.g., progressive P-frame or B-frame, interlaced P-field or B-field, or interlaced P-frame or B-frame) is represented in terms of prediction from one or more other pictures (which are typically referred to as reference pictures or anchors). A prediction residual is the difference between predicted information and corresponding original information. In contrast, a key picture (e.g., progressive I-frame, interlaced I-field, or interlaced I-frame) is compressed without reference to other pictures.

If the current picture (405) is a predicted picture, a motion estimator (410) estimates motion of macroblocks or other sets of samples of the current picture (405) with respect to one or more reference pictures. The picture store (420) buffers a reconstructed previous picture (425) for use as a reference picture. When multiple reference pictures are used, the multiple reference pictures can be from different temporal directions or the same temporal direction. The motion estimator (410) outputs as side information motion information (415) such as differential motion vector information.

The motion compensator (430) applies reconstructed motion vectors to the reconstructed (reference) picture(s) (425) when forming a motion-compensated current picture (435). The difference (if any) between a block of the motion-compensated current picture (435) and corresponding block of the original current picture (405) is the prediction residual (445) for the block. During later reconstruction of the current picture, reconstructed prediction residuals are added to the motion-compensated current picture (435) to obtain a reconstructed picture that is closer to the original current picture (405). In lossy compression, however, some information is still lost from the original current picture (405). Alternatively, a motion estimator and motion compensator apply another type of motion estimation/compensation.

A frequency transformer (460) converts spatial domain video information into frequency domain (i.e., spectral, transform) data. For block-based video pictures, the frequency transformer (460) applies a DCT, variant of DCT, or other forward block transform to blocks of the samples or prediction residual data, producing blocks of frequency transform coefficients. Alternatively, the frequency transformer (460) applies another conventional frequency transform such as a Fourier transform or uses wavelet or sub-band analysis. The frequency transformer (460) may apply an 8×8, 8×4, 4×8, 4×4 or other size frequency transform.

A quantizer (470) then quantizes the blocks of transform coefficients. The quantizer (470) applies non-uniform, scalar quantization to the spectral data with a step size that varies on a picture-by-picture basis or other basis. The quantizer (470) can also apply another type of quantization to the spectral data coefficients, for example, a uniform or adaptive quantization for at least some of the coefficients, or directly quantizes spatial domain data in an encoder system that does not use frequency transformations. In described embodiments, the quantizer (470) (in conjunction with other modules such as a rate controller) controls encoding quality for textured, dark smooth and other smooth video content by adjusting quantization step size and/or DZ size.

When a reconstructed current picture is needed for subsequent motion estimation/compensation, an inverse quantizer (476) performs inverse quantization on the quantized spectral data coefficients. An inverse frequency transformer (466) performs an inverse frequency transform, producing blocks of reconstructed prediction residuals (for a predicted picture) or samples (for a key picture). If the current picture (405) was a key picture, the reconstructed key picture is taken as the reconstructed current picture (not shown). If the current picture (405) was a predicted picture, the reconstructed prediction residuals are added to the motion-compensated predictors (435) to form the reconstructed current picture. One or both of the picture stores (420, 422) buffers the reconstructed current picture for use in subsequent motion-compensated prediction.

The entropy coder (480) compresses the output of the quantizer (470) as well as certain side information (e.g., motion information (415), quantization step size). Typical entropy coding techniques include arithmetic coding, differential coding, Huffman coding, run length coding, LZ coding, dictionary coding, and combinations of the above. The entropy coder (480) typically uses different coding techniques for different kinds of information, and can choose from among multiple code tables within a particular coding technique.

The entropy coder (480) provides compressed video information (495) to the multiplexer (“MUX”) (490). The MUX (490) may include a buffer, and a buffer level indicator may be fed back to a controller. Before or after the MUX (490), the compressed video information (495) can be channel coded for transmission over the network.

A controller (not shown) receives inputs from various modules such as the motion estimator (410), frequency transformer (460), quantizer (470), inverse quantizer (476), entropy coder (480), and buffer (490). The controller evaluates intermediate results during encoding, for example, setting quantization step sizes and performing rate-distortion analysis. The controller works with modules such as the motion estimator (410), frequency transformer (460), quantizer (470), and entropy coder (480) to classify types of content, and to set and change coding parameters during encoding. When an encoder evaluates different coding parameter choices during encoding, the encoder may iteratively perform certain stages (e.g., quantization and inverse quantization) to evaluate different parameter settings. The encoder may set parameters at one stage before proceeding to the next stage. Or, the encoder may jointly evaluate different coding parameters. The tree of coding parameter decisions to be evaluated, and the timing of corresponding encoding, depends on implementation.

The relationships shown between modules within the encoder (400) indicate general flows of information in the encoder; other relationships are not shown for the sake of simplicity. In particular, FIG. 4 usually does not show side information indicating the encoder settings, modes, tables, etc. used for a video sequence, picture, macroblock, block, etc. Such side information, once finalized, is sent in the output bitstream, typically after entropy encoding of the side information.

Particular embodiments of video encoders use a variation or supplemented version of the generalized encoder (400). Depending on implementation and the type of compression desired, modules of the encoder can be added, omitted, split into multiple modules, combined with other modules, and/or replaced with like modules. For example, the controller can be split into multiple controller modules associated with different modules of the encoder. In alternative embodiments, encoders with different modules and/or other configurations of modules perform one or more of the described techniques.

III. Distinguishing Types of Picture Content

The human visual system perceives distortion differently in textured picture content, dark smooth picture content, and other smooth picture content. Generally, a given amount of distortion is more noticeable and distracting in non-textured content than textured content. Among non-textured content, a given amount of distortion is usually more noticeable and distracting in dark smooth content than other smooth content. Classifying picture content as textured, smooth or dark smooth facilitates allocation of available bits to the different categories of content, so as to control the perceptibility of distortion when distortion is introduced.

In some embodiments, a tool such as a video encoder classifies blocks of sample values in video pictures into three categories: smooth, dark smooth, and textured. The texture of a block generally corresponds to amount of detail in the block. It can also relate to the amount of energy in different frequency coefficients for the block. A block with complex, intricate detail has high texture, and higher frequency AC coefficients for the block usually represent much of the energy for the block. In contrast, a “smooth” block can include uniform or nearly uniform sample values or, more generally, content classified as non-textured according to a texture metric in use. Smooth regions can include flat regions (areas of constant or nearly constant color) or gradient slope regions (areas of color that vary at a constant or nearly constant rate across the region), and smooth regions may be considered smooth even when interrupted by small areas of noise, film grains, or other color variations. A “dark smooth” block, generally, includes content classified as dark according to a dark content threshold or range in use.

An effective threshold between textured and smooth video content can vary depending on the type of content being encoded. For example, a threshold that effectively separates textured content from non-textured content in natural video may classify all blocks in animation video as non-textured, which hampers effective control of distortion in the animation video. In practice, an effective smooth/textured threshold for animation content is usually lower than an effective smooth/textured threshold for natural video, and the gradations between effectively different thresholds are finer for animation content.

A. Distinguishing Textured Content from Non-textured Content

In some embodiments, a tool such as a video encoder distinguishes textured content from non-textured content using a non-linear scale of texture thresholds. The non-linear scale embodies finer gradations of possible texture threshold values in some ranges, and embodies coarser gradations of possible texture threshold values in other ranges, which can help control encoding for different categories of content. Animation content usually includes relatively uniform regions of sample values, whereas natural video usually includes subtle or complex variations in sample values. In many encoding scenarios, coarse gradations of texture threshold values provide effective control over the smooth/textured classification for natural video, as different thresholds allow the user to designate slightly different proportions of content as textured. Such coarse gradations may be ineffective for animation content, however, if one threshold value results in the classification of most of the content as textured, and the next threshold value results in the classification of most of the content as non-textured, with no way to classify an intermediate proportion of content as textured. Conversely, threshold gradations that work effectively for animation content may be too fine for natural video, such that large changes in threshold value fail to change the proportion of content classified as textured. Using a scale of texture thresholds with non-linear gradations helps improve control of smooth/textured threshold classification for different categories of video content.

FIG. 5 shows a generalized technique (500) for using texture thresholds separated by non-linear gradations when classifying content. A tool such as the encoder (400) explained with reference to FIG. 4 or other tool performs the technique (500) as part of encoding or in another process.

First, the tool gets (510) a level of texture classification. For example, the tool gets (510) a level of texture classification specified by the user through a graphical user interface control of the tool. Using a slider bar or simple numerical scale, the user can set different levels of texture classification in an intuitive way, so as to control the proportion of content classified as textured versus non-textured. In a typical configuration, setting a lower level of texture classification results in more of the content being classified as textured, and setting a higher level results in less of the content being classified as textured. Alternatively, the tool gets (510) the level of texture classification through an encoding wizard, command line interface, or other way, or uses a default level of texture classification.

The tool then sets (520) a texture threshold using a non-linear mapping of possible values of the texture classification level to possible values of the texture threshold. According to an example non-linear mapping, lower levels of texture classification map to lower texture thresholds separated by finer gradations, and higher levels map to higher texture thresholds separated by coarser gradations. As a result, for relatively simple content such as animation content, finer gradations at low texture thresholds provide for effective differentiation of proportions of textured versus non-textured content. On the other hand, for more complex content such as natural video, the coarser gradations provide effective smooth/textured differentiation at higher texture thresholds. Alternatively, the non-linear mapping embodies other and/or additional rules. Generally, the tool sets (520) the texture threshold using a scale with non-linear gradations between possible values of the texture threshold. The non-linear gradations include finer gradations for some ranges and coarser gradations for other ranges, as desired according to a particular implementation.

The tool classifies (530) multiple units (e.g., blocks) of sample values using the texture threshold. For example, for each of the multiple units, the tool compares a texture metric to the texture threshold that was set. In an example implementation, the texture metric for a block is the sum of squared AC coefficients for the block, and the texture threshold is an AC energy threshold. Alternatively, the texture metric measures sum of differences between intensity values of a block and their average value, intensity variance (e.g., using sums of difference between adjacent sample values, or using an average gradient value, or using a median gradient value), and the threshold varies accordingly. For examples of alternative texture metrics, see U.S. Patent Application Publication No. US-2007-0248164-A1, entitled, “QUANTIZATION ADJUSTMENT BASED ON TEXTURE LEVEL.” Alternatively, the tool uses another texture metric and corresponding texture threshold.

In the example implementation, the tool classifies macroblocks as textured or non-textured (pending further differentiation as dark smooth or other smooth) based on the classifications of their constituent blocks. A macroblock in 4:2:0 chroma sampling format includes four luma blocks and two chroma blocks. If any of the six blocks is classified as smooth/non-textured, the macroblock is classified as smooth/non-textured. Alternatively, the tool uses a different rule for classifying macroblocks, such as a rule based on whether or not a majority of the blocks in a macroblock are classified as smooth.

Although FIG. 5 shows the classifying (530) immediately following the setting (520) of the texture threshold, the tool set (520) the texture threshold on a session-by-session basis for encoding sessions. Alternatively, the tool sets the texture threshold during real-time encoding, from picture-to-picture during offline studio editing and encoding, or on some other basis. The tool can perform the classifying (530) on a picture-by-picture basis or other basis.

For an example mapping, the user of an encoding system selects a level l of texture classification, which the encoding system maps to a texture threshold t according to the relation:

t=4096*(l/MAXLEVEL) custom character N (2).

MAXLEVEL indicates the maximum allowable level of texture classification, and N indicates a scaling factor. The values of MAXLEVEL and N depend on implementation. When N=1, the mapping of level l to threshold t is linear.

To speed up processing, the encoding system can include a table that associates level values with threshold values. The following table shows an approximate non-linear mapping of possible values of texture classification level l to possible values of texture threshold t when the scaling factor N is 1.5, and the maximum level value MAXLEVEL is 100.

level l
threshold t
gap

1
4

2
12
7

3
21
10

4
33
11

5
46
13

6
60
14

7
76
16

8
93
17

9
111
18

10
130
19

11
149
20

12
170
21

13
192
22

14
215
23

15
238
23

16
262
24

17
287
25

18
313
26

19
339
26

20
366
27

21
394
28

22
423
28

23
452
29

24
482
30

25
512
30

26
543
31

27
575
32

28
607
32

29
640
33

30
673
33

31
707
34

32
741
34

33
776
35

34
812
36

35
848
36

36
885
37

37
922
37

38
959
38

39
998
38

40
1036
39

41
1075
39

42
1115
40

43
1155
40

44
1195
41

45
1236
41

46
1278
41

47
1320
42

48
1362
42

49
1405
43

50
1448
43

51
1492
44

52
1536
44

53
1580
45

54
1625
45

55
1671
45

56
1716
46

57
1763
46

58
1809
47

59
1856
47

60
1904
47

61
1951
48

62
2000
48

63
2048
49

64
2097
49

65
2146
49

66
2196
50

67
2246
50

68
2297
50

69
2348
51

70
2399
51

71
2450
52

72
2502
52

73
2555
52

74
2607
53

75
2660
53

76
2714
53

77
2768
54

78
2822
54

79
2876
54

80
2931
55

81
2986
55

82
3041
55

83
3097
56

84
3153
56

85
3210
56

86
3267
57

87
3324
57

88
3381
57

89
3439
58

90
3497
58

91
3556
58

92
3614
59

93
3674
59

94
3733
59

95
3793
60

96
3853
60

97
3913
60

98
3974
61

99
4036
61

100
4096
61

Alternatively, the tool distinguishes textured content from non-textured content using another approach. For example, the tool uses a static texture threshold or uses an adjustable texture threshold according to a scale with uniform gradations between possible values of the texture threshold.

B. Distinguishing Smooth Content from Dark Smooth Content

In some embodiments, a tool such as a video encoder distinguishes dark smooth content from other smooth content. This can help the tool control encoding quality of the dark smooth content relative to encoding quality of the other smooth content. Some types of distortion are more perceptible in dark smooth content than in other smooth content. Allocating bits in same way across all types of smooth content can result in perceptible distortion in dark smooth areas. Classifying picture content as smooth or dark smooth can facilitate allocation of available bits to the dark smooth content and other smooth content, so as to balance or otherwise control the perceptibility of distortion when distortion is introduced.

FIG. 6 shows a generalized technique (600) for classifying units (e.g., blocks) of sample values as smooth or dark smooth. A tool such as the encoder (400) explained with reference to FIG. 4 or other tool performs the technique (600) as part of encoding or in another process.

First, the tool gets (610) a level of dark smooth content classification. For example, the tool gets (610) a default level of dark smooth content classification or gets (610) a level of dark smooth content classification specified by a user through a graphical user interface control of the tool. If user control is desirable, the user can set different levels of dark smooth content classification to control the proportion of content classified as dark smooth versus smooth. In a typical configuration, setting a lower level results in less content being classified as dark smooth, and setting a higher level results in more of the content being classified as dark smooth. Alternatively, the tool gets (610) the level of dark smooth content classification through an encoding wizard, command line interface, or other way.

The tool sets (620) one or more ranges of dark smooth content values using the level of dark smooth content classification. Typically, for a conservative level, the range of dark smooth content values is smaller. Or, for a more aggressive level of dark smooth content classification, the range is larger and more content qualifies as dark smooth content.

The tool classifies (630) multiple units (e.g., blocks) of sample values using the range(s) for dark smooth content classification. In the example implementation, for each of the multiple units, the tool compares a metric to one of the ranges that was set. The metric for a block is the average intensity value or DC coefficient value for the block, and the range is a range of intensity values or DC coefficient values. For example, for default designation of content as dark smooth content, average luma intensity values between 16 and 40 indicate dark smooth content, and average chroma sample values are between 120 and 136 indicate dark smooth content. Or, for less aggressive designation of content as dark smooth content, the ranges are 16 to 28 for luma and 124 to 132 for chroma. Alternatively, the tool uses another darkness metric and corresponding ranges for dark smooth content.

In the example implementation, the tool classifies a macroblock as smooth/non-textured whether it includes dark smooth content or other smooth content, and uses other control parameters to regulate relative encoding quality of dark smooth content versus other smooth content. Alternatively, the tool classifies macroblocks as smooth or dark smooth based on the classifications of their constituent blocks. According to one alternative rule, if any of the blocks in a macroblock is classified as dark smooth, the macroblock is classified as dark smooth. Or, the tool uses a different rule for classifying macroblocks based on whether the macroblock includes more dark smooth blocks than other smooth blocks.

Although FIG. 6 shows the classifying (630) immediately following the setting (620) of the dark smooth content ranges, the tool can set (620) the ranges on a session-by-session basis for encoding sessions or use default ranges. Alternatively, the tool sets the ranges during real-time encoding, from picture-to-picture during offline studio editing and encoding (automatically distinguishing areas of dark smooth content can simplify and dramatically speed up the process of removing perceptible encoding artifacts in dark smooth areas), or on some other basis. The tool can perform the classifying (630) on a picture-by-picture basis or other basis.

C. Block-by-Block Classification as Textured, Dark Smooth or Smooth

FIG. 7 shows an example technique (700) for classifying blocks of video pictures as textured, smooth or dark smooth. A tool such as the encoder (400) explained with reference to FIG. 4 or other tool performs the technique (700) as part of encoding or in another process.

FIG. 7 shows classification of blocks in a picture on a block-by-block basis. To start, the tool gets (710) the next block and applies (720) a frequency transform to the sample values of the block. For example, the tool gets an 8×8 block and applies a frequency transform (such as a discrete cosine transform or similar transform) to the block to produce an 8×8 block of transform coefficients. Alternatively, the tool gets a block of a different size (e.g., 8×4, 4×8 or 4×4) and/or applies a different type of transform. In the 8×8 matrix of transform coefficients, the horizontal and vertical positions of coefficients correspond to horizontal and vertical frequencies, respectively, for the signal represented in the samples values of the 8×8 block. The coefficient shown by convention as the upper left coefficient (or (0, 0) coefficient) corresponds to the lowest frequency in both the vertical and horizontal directions, and is termed the DC coefficient for the block. The other coefficients are termed AC coefficients for the block. The AC coefficient shown by convention as the lower right coefficient (or (7, 7) coefficient) represents the highest horizontal and vertical frequency component for the 8×8 block.

The tool measures (730) the energy of the AC coefficients for the block and compares (740) the measured AC energy to a texture threshold. In an example implementation, the tool computes the sum of squares of the AC coefficients and compares the sum to a texture threshold. Alternatively, the tool computes another measure of AC energy or uses another texture metric, varying the texture threshold in use accordingly.

If the measured energy is above the texture threshold, the tool classifies (750) the block as textured. For blocks in a picture, the tool uses a higher texture threshold for luma blocks than for chroma blocks, since there is typically more AC energy in the luma plane than the chroma plane.

Otherwise, the block is deemed smooth (not textured) and the tool checks (760) whether the DC coefficient value of the block is within the range associated with dark smooth content. If so, the tool classifies (780) the block as dark smooth. Otherwise, the tool classifies (770) the block as just smooth. In the example implementation, if a chroma block is not textured, the tool classifies the chroma block as smooth or dark smooth depending on the DC coefficient values of the corresponding luma block(s) for the chroma block. If a majority of the corresponding luma blocks have DC coefficients in the dark smooth range, the tool classifies the chroma block as dark smooth, and otherwise the tool just classifies the chroma block as smooth. Alternatively, the tool uses another rule to classify chroma blocks as smooth or dark smooth depending on luma block classifications. Or, chroma blocks are individually classified, and the tool checks DC coefficients of chroma blocks against dark smooth content ranges.

After the tool classifies the block, the tool determines (790) whether to continue by getting (710) the next block. The encoder repeats the process on a block-by-block basis (as shown), macroblock-by-macroblock basis or other basis. Also, for the sake of simplicity, FIG. 7 does not show how the technique (700) interacts with other encoding processes such as frequency transformation, entropy encoding and rate control.

In the example implementation, the tool subsequently processes macroblocks or other areas of the picture using the block classifications. For example, the tool classifies a macroblock as textured or non-textured. If a macroblock is non-textured, the blocks of the macroblock can be further classified as smooth or dark smooth and thereby designated for special treatment in quantization. The tool uses other control parameters to process smooth blocks and dark smooth blocks differently within smooth/non-textured macroblocks. Alternatively, the tool directly classifies macroblocks as textured, dark smooth, or smooth.

The tool can record results of the classification using a map that indicates texture levels or otherwise identifies textured units (e.g., blocks, macroblocks) and non-textured units (e.g., blocks, macroblocks). Later processes can then use the map. Alternatively, later processes directly perform the classification.

IV. Adjusting Encoding Quality for Textured, Dark Smooth and Other Smooth Content

In some embodiments, an encoder adjusts encoding quality for textured, dark smooth, and other smooth video content. A goal of the bit allocation techniques and tools described herein is to introduce distortion where it is less perceptible when distortion has to be introduced during encoding to satisfy encoding constraints.

A. Generalized Adjustment Techniques

FIG. 8 shows a generalized technique (800) for allocating bits between textured video content, dark smooth video content and other smooth video content. A tool such as the encoder (400) explained with reference to FIG. 4 or other tool performs the technique (800) as part of encoding.

FIG. 8 shows adjustment of encoding quality for a single unit of frequency coefficients. The technique (800) includes one path for control bit allocation in textured units, and another path for controlling bit allocation in smooth and dark smooth units. For the sake of simplicity, FIG. 8 does not show how the technique (800) interacts with other encoding processes such as frequency transformation and rate control.

The encoder determines (810) whether the unit being encoded is textured. If so, the encoder quantizes (820) the frequency coefficients of the unit using quantization for a textured unit. In an example implementation, for a textured macroblock, the encoder uses the quantization parameter (“QP”) value at picture-level for the picture that includes the textured macroblock to quantize the frequency coefficients of the blocks of the macroblock. In doing so, the encoder can adapt the DZ size depending on the amount of texture in the unit. Alternatively, the encoder uses another approach to quantizing the textured unit. For example, see U.S. Patent Application Publication No. US-2007-0248164-A1.

If the unit is non-textured, the encoder selectively quantizes (830) the frequency coefficients of the unit using quantization for a dark smooth unit and/or using quantization for a smooth unit. In the example implementation, the encoder finds and uses a macroblock-level QP value for a non-textured macroblock to quantize the frequency coefficients of the blocks of the macroblock. On a block-by block basis, the encoder uses one or more control parameters (such as the TACL, TACC, TACLD and TACCD parameters described below) to regulate quantization of frequency coefficients in dark smooth blocks and smooth blocks. The encoder uses the macroblock-level QP for quantization of the respective blocks, adjusting DZ size from block to block. Alternatively, the encoder uses another approach to quantizing the non-textured unit. For example, if a unit (e.g., macroblock) is classified as smooth or dark smooth, the encoder determines whether the unit is classified as dark smooth. If so, the encoder quantizes the frequency coefficients of the unit using a QP and DZ size adapted to the dark smooth unit. Otherwise, the encoder quantizes the frequency coefficients of the unit using a QP and DZ size for a smooth unit.

The encoder repeats the technique (800) on a unit-by-unit basis or on some other basis. The encoder can use macroblock-level classifications for some decisions (e.g., smooth or textured) and use block-level classifications for other decisions (e.g., dark smooth or other smooth).

B. Example Implementation of Encoding Adjustments

Before detailing exemplary ways for encoders to adjust encoding quality for textured, dark smooth, and other smooth video content, this section describes quantization and inverse quantization rules in an example implementation.

1. Example QPs, Reconstruction Rules, and Quantization Rules

In an example implementation, a decoder (or encoder during encoder-side reconstruction) uses either of two different inverse quantization modes (also called reconstruction rules) when reconstructing AC coefficients for a block. The decoder or encoder uses a “uniform” quantizer reconstruction rule or a “non-uniform” quantizer reconstruction rule. The mode is specified for a picture, and all blocks in the picture use the same mode. The value of QP can vary within a picture for some encoding settings, and can take any integer value between 1 and 31 inclusive. The QP value is multiplied by a factor of 2 when used in inverse quantization. (The quantization factor can also be modified by a “half QP” step, but for the sake of simplicity this halfQP step quantization is typically not addressed in the following discussion.)

According to the “uniform” quantizer reconstruction rule, the decoder or encoder reconstructs AC coefficients as follows:

C=2×QP×L (3),

where C is the reconstructed, inverse quantized AC coefficient, and L is the quantized level for the coefficient. Reconstruction points appear as circles in FIG. 9.

The quantization rules used to convert the AC coefficients to quantized levels L are not specified by the reconstruction rule (3). Rather, quantization rules, including placement of the quantization bin boundaries, are defined by the encoder. Suppose the distribution of AC coefficient values was uniform. For the “uniform” quantizer reconstruction rule, the corresponding quantization rule that would produce minimum distortion would be:

L=sign(C)×((abs(C)+QP)//(2×QP)) (4),

where sign(C)=−1 for C<0, and sign(C)=1 for C>=0. The // operation indicates integer division with rounding towards zero, and abs(C) indicates the absolute value of C. For many real-world encoding scenarios, however, the quantization rule shown in (4) does not provide optimal rate-distortion performance. For one thing, the distribution of values within quantization bins is typically not normal. Moreover, the quantization rules do not account for differences in the perceptibility of distortion in different situations. Thus, many encoders use different quantization rules, which vary depending on implementation.

Example quantization bin boundaries appear as dashed lines in FIG. 9. Compared to theoretical uniform quantization bins (which would appear at mid-points between reconstruction points), the boundaries shown in the top row of FIG. 9 are slightly further away from the zero-value point. In practice, the quantization bin boundaries used during quantization need not have been at the positions shown in FIG. 9. In the lower two rows of FIG. 9, the DZ is widened, without changing the other quantization thresholds.

According to the “non-uniform” quantizer reconstruction rule, the decoder or encoder reconstructs non-zero AC coefficients as follows:

C=(2×L+1)×QP (5).

Reconstruction points appear as circles in FIG. 10. In this mode, even for default settings, the quantization bin that includes the zero-value reconstruction point (the DZ) is much wider than the others. For the “non-uniform” quantizer reconstruction rule, a simple corresponding quantization rule would be:

L=sign(C)×(abs(C)//(2×QP)) (6).

Example quantization bin boundaries appear as dashed lines in FIG. 10. In the lower two rows of FIG. 10, the DZ is widened without modifying the other quantization thresholds.

Alternatively, the decoder/encoder uses more/fewer rules and/or different rules during reconstruction, and the encoder uses more/fewer rules and/or different rules during quantization. Generally, the parameterization of quantization parameters, quantization rules, and reconstruction rules can vary from implementation to implementation.

2. Example Approaches to Preserving AC Coefficients of a Block

In the example implementation, an encoder selectively preserves AC coefficients that have the most energy in a smooth or dark smooth block, so as to preserve encoding quality. The encoder selects a QP and DZ for the block such that N significant AC coefficients are preserved after quantization and inverse quantization, where N is an integer greater than or equal to 1 and less than the total number of AC coefficients for the block. Often, the preserved AC coefficients are coefficients for the lowest frequency AC basis functions of the transform, which characterize gradual value changes horizontally and/or vertically across a block. Preserving these AC coefficients tends to help perceived visual quality for the block, especially for smooth regions with low variation. The encoder selects a QP that preserves the desired number of the top-amplitude AC coefficients at acceptable quality. The N AC coefficients are preserved with a reasonable QP—typically not QP=1, which might result in allocation of more bits than necessary for the N AC coefficients. There can be situations (e.g., very flat blocks that lack N significant AC coefficients) in which the target number of AC coefficients is not preserved. In general, however, the encoder preserves AC coefficients without being overly aggressive in spending bits with smaller QPs.

In some cases, coefficients after the top N AC coefficients have values close to the N^thAC coefficient. The encoder can reduce bit rate by adjusting the DZ threshold used for the given QP to quantize coefficients after the top NAC coefficients to zero. In general, DZ(QP) indicates a “cut-off” threshold for quantizing an AC coefficient to zero when the value of QP is used for quantization. For example, if the DZ range is −200 to +200 then a coefficient that is 197 would be quantized to L=0, and a coefficient that is 202 would be quantized to L=1.

In some encoders, the default DZ threshold DZ(QP) for a given QP is predetermined and proportional to the QP. Instead of using predetermined DZ thresholds when preserving N AC coefficients, the encoder can more aggressively set a DZ threshold. FIG. 9 shows three different cut-off thresholds for the uniform quantizer mode, one at the default placement of |6*QP/5|, one closer to the first reconstruction point, and one after the first reconstruction point. FIG. 10 shows three different cut-off thresholds for the non-uniform quantizer mode. In these examples, increasing the size of the DZ does not push the other quantizer thresholds further away from zero; it just results in more numbers close to zero being quantized to zero.

In the example implementation, the encoder adaptively adjusts the DZ size of a quantizer so as to preserve the N most significant AC coefficients in a block, which causes AC coefficients smaller than the N^thlargest AC coefficient (in absolute value) to be quantized to zero, even if they would otherwise be preserved using the normal DZ threshold for the selected QP. If the normal DZ threshold is |6*QP/5| for uniform quantization, for example, the encoder adjusts DZ threshold to be the absolute value of the N^thmost significant AC coefficient. This helps the video encoder selectively reduce bit rate, making bits available for use elsewhere in encoding.

FIG. 11 shows a block of frequency coefficients in which the DC coefficient and lower frequency AC coefficients have most of the signal energy for the block. In this example, the transform coefficients are in 12.4 format, meaning that the 12 most significant bits represent the integer portion of a coefficient, and the 4 least significant bits represent the fractional portion of the coefficient. The DC coefficient has a value of 296, and the most significant AC coefficients have values of −867, 311, −241, 202, 197, 188 and 105. The encoder selects a QP and DZ size to preserve a target number of AC coefficients. If N=4, the encoder selects a QP and DZ according to which the AC coefficients −867, 311, 241 and 202 are preserved. One option is to set QP=1 and use a relatively wide DZ to quantize the coefficients less than 202 to zero. While this preserves the four most significant coefficients, encoding these four coefficients can consume an inefficient number of bits. FIGS. 12A and 12B illustrate other options.

FIG. 12A illustrates more aggressive quantization that still preserves the top four AC coefficients in the block shown in FIG. 11. The encoder uses QP=6, for which the normal DZ threshold is 6*QP*16/5=115 (uniform quantizer mode). Using QP=6 and the default DZ preserves the four most significant coefficients (−867, 311, −241 and 202), but two other AC coefficients are also preserved (197 and 188). This reduces the bit rate associated with encoding the four most significant coefficients (QP=6 vs. QP=1) but includes AC coefficients over the target number.

FIG. 12B illustrates the aggressive quantization of FIG. 12A (QP=6) with further adjustment of the DZ to quantize to zero any AC coefficients after the four most significant AC coefficients. Two AC coefficients (188, 197) whose values are close to the 4^thmost significant coefficient (202) are quantized to zero. This reduces the bit rate associated with encoding the four most significant coefficients (QP=6 vs. QP=1) and also limits the encoded AC coefficients to the target number.

In many cases, given the adjustment of the DZ to limit encoding to the target number of coefficients, a range of values of QP can provide acceptable performance. In the example shown in FIG. 11, QP values of 4, 5, 7 or 8 could be used to encode the four most significant AC coefficients with suitable quality.

In the example implementation, the encoder jointly determines the DZ threshold and QP as follows. The encoder sets the DZ size based upon the N^thmost significant AC coefficient. The encoder then computes a quantization step size as int(DZ/16), where int(x) returns the integer portion of the real number x, and the division by 16 accounts for the fractional component of the 12.4 format for AC coefficient values. The quantization step size is parameterized as 2*QP, and the encoder determines QP (and potentially halfQP) from the quantization step size.

For example, the encoder sets the DZ(QP) to be equal to the absolute value of the N^thlargest AC coefficient among the 63 AC coefficients for an 8×8 block. If N=2 and the 2^ndlargest AC coefficient (in terms of absolute value) is −643, the encoder sets the DZ range to be −643 to +643. The two most significant AC coefficients are quantized to non-zero values, and the rest of the AC coefficients for the block are quantized to zero. The encoder then computes the quantization step size as int(659/16)=40, and determines that QP=20. If the N^thlargest AC coefficient can be preserved using a value of QP that exceeds the allowable range, the maximum value for QP can be used, or the QP value used for the picture including the block can be used.

3. Example Adjustments for Smooth and Dark Smooth Content

In the example implementation, an encoder selectively preserves AC coefficients in dark smooth and smooth units so as to preserve encoding quality for the AC coefficients. For each macroblock with one or more smooth or dark smooth blocks, the encoder selects a QP and DZ(s) such that there are at least N non-zero quantized AC coefficients per block of the macroblock.

When selecting the N most significant AC coefficients per block, the parameters that an encoder uses depend on implementation, and can help control bit allocation between dark smooth blocks and other smooth blocks. In the example implementation, the encoder uses a target number of AC coefficients for luma blocks of other smooth content (“TACL”), a target number of AC coefficients for chroma blocks of other smooth content (“TACC”), a target number of AC coefficients for luma blocks of dark smooth content (“TACLD”), and a target number of AC coefficients for chroma blocks of dark smooth content (“TACCD”). Alternatively, the encoder uses a different parameterization of target number(s) of AC coefficients for dark smooth blocks and other smooth blocks. For example, N is the same for chroma blocks and luma blocks of the macroblock, and N is the same for dark smooth blocks and other smooth blocks. Or, N is different for chroma blocks and luma blocks, but does not change depending on whether a block is dark smooth or other smooth. Or, N is different for dark smooth blocks and other smooth blocks, but does not change depending on whether a block is luma or chroma.

Depending on implementation, an encoder can change TACL, TACC, TACLD, and TACCD on a picture-by-picture basis or on some other basis, using feedback from a rate controller and/or user settings. The encoder can use a single target number of AC coefficients (“TAC”) for a picture, with TACL, TACC, TACLD, and TACCD in turn depending on the picture-level value TAC. One possible configuration is TACL=TAC, TACC=int(TAC/2), TACLD=TAC+2, and TACCD=int((TAC+2)/2). Other configurations use other relations between the control parameters.

FIG. 13 shows an example technique (1300) for determining QP and DZ size(s) for macroblocks that include dark smooth blocks or other smooth blocks. A tool such as the encoder (400) explained with reference to FIG. 4 or other tool performs the technique (1300) as part of encoding.

FIG. 13 shows determination of QP and DZ size for a single macroblock. The technique (1300) shown in FIG. 13 can be used in a way that integrates quantization and rate control. For the sake of simplicity, however, FIG. 13 does not show how the technique (1300) interacts with other encoding processes such as frequency transformation, texture/dark smooth/smooth classification, and rate control.

First, the encoder gets (1310) the N^thmost significant AC coefficient for each dark smooth or other smooth block in the macroblock. The encoder can use a texture map or other record of block classifications, or compute block classifications at this time. So, if an encoder using TACL, TACC, TACLD, and TACCD encodes a macroblock with two dark smooth luma blocks, two other smooth luma blocks, and two dark smooth chroma blocks, the encoder uses the appropriate one of TACL, TACLD, or TACCD to get the N^thmost significant AC coefficient value for each of the blocks in the macroblock, where TACL, TACLD, and TACCD potentially yield different values of N.

The encoder then finds (1340) a value for QP for the macroblock. The encoder finds a value for QP that preserves the N most significant coefficients for each dark smooth or other smooth block in the macroblock. So, if an encoder uses TACL, TACC, TACLD, and TACCD, the encoder finds a QP that satisfies the applicable one of TACL, TACC, TACLD, and TACCD for every block in the macroblock, depending on whether the respective blocks are dark smooth, other smooth, or textured, and depending on whether the respective blocks are luma or chroma. For example, the encoder determines the lowest amplitude N^thAC coefficient (min_Nth_AC) among the dark smooth blocks and other smooth blocks of the macroblock. If Y_block1_AC(TACLD), Y_block2_AC(TACLD), Y_block3_AC(TACL), Y_block4_AC(TACL), U_block_AC(TACCD), and V_block_AC(TACCD) represent the N^thAC coefficients for a set of dark smooth and other smooth blocks of a macroblocks, respectively, min_Nth_AC=minimum (Y_block1_A C(TACLD), Y_block2_AC(TACLD), Y_block3_AC(TACL), Y_block4_AC(TA CL), U_block_AC(TACCD), and V_block_AC(TACCD)). The encoder then determines a QP from min_Nth_AC using a mechanism described in the previous section. If a quantization step size based on QP preserves min_Nth_AC, it also preserves the N^thAC coefficient for the respective other blocks in the macroblock. Thus, the encoder finds a QP for the macroblock directly from the smallest of the N^thAC coefficients of any dark smooth blocks or smooth blocks in the macroblock.

Alternatively, to find (1340) the QP for a macroblock, the encoder compares min_Nth_AC with the default DZ(QP) for candidate values of QP. The encoder starts with the picture QP and decreases QP until a minimum QP for the quantizer is reached (e.g., QP=1) or the condition |min_Nth_AC|>=DZ(QP) is satisfied for the default DZ(QP). If the condition is satisfied, the encoder sets the threshold QP for the macroblock to be the first QP (i.e., highest qualifying QP) that satisfies the condition.

Next, the encoder quantizes the blocks of the macroblock using the QP for the macroblock. The encoder gets (1350) the next block to quantize in the macroblock and determines (1360) whether the block is textured or non-textured. If the block is textured, the encoder quantizes (1370) the block using the picture-level QP and the default DZ threshold for QP. Alternatively, the encoder adapts the DZ threshold used depending on the texture content of the textured block.

If the block is not textured, the encoder quantizes (1380) the block using the macroblock-level QP and adjusting the DZ threshold. In doing so, the encoder can increase the DZ threshold for the block to quantize to zero any coefficients after the N^thAC coefficient in the block, where N depends on TA CL, TA CC, TACLD, or TACCD, or is otherwise set by the encoder. The encoder then determines (1390) whether to continue by getting (1350) the next block in the macroblock or ending. Thus, for each smooth or dark-smooth block, the encoder determines the DZ threshold that preserves target number of AC coefficients, for example, using the appropriate TAC parameter.

4. Signaling Quantization Parameters

In the example implementation, the encoder signal macroblock-level QP values for non-textured macroblocks using differential quantization relative to picture-level QP values. The encoder quantizes textured blocks in a picture using the picture-level QP value, and it signals the picture-level QP value and reconstruction rule for the picture as part of a picture header. For a non-textured macroblock in a picture, the encoder signals the macroblock-level QP value for the macroblock as part of a macroblock header. (The picture-level reconstruction rule is used for the non-textured macroblocks in the picture.) The macroblock-level QP value can be encoded as a difference (e.g., a reduction in QP value) relative to the applicable picture-level QP value.

Alternatively, the encoder uses another mechanism to signal quantization parameters for textured content and non-textured content.

V. Setting Control Parameters for Bit Allocation

In some embodiments, an encoder sets bit allocations for adjusting encoding quality of textured content, dark smooth content, and other smooth content using a multi-pass approach. In an example implementation, the encoder uses QP, TACL, TACC, TACLD, and TACCD as control parameters for bit allocations. Alternatively, the encoder uses other control parameters.

In the example implementation, an encoder sets bit allocations in peak-bit-rate-constrained, variable-bit-rate encoding scenarios. When encoding a video sequence, the encoder sets an average bit rate target (e.g., corresponding to an average network connection rate or average transmission rate), a peak bit rate (e.g., corresponding to a peak network connection rate or peak transmission rate), and a buffer size (e.g., corresponding to a notional encoder output buffer size or actual decoder input buffer size). The encoder controls encoding so that the resulting bit stream of compressed video information conforms to the average bit rate and the buffer level does not underflow. Stated differently, the encoder controls encoding so that the total size of the encoded video (indicated by the target bit rate x the duration of the video sequence) matches a target size. The buffer level drops when the current bit rate exceeds the peak bit rate (since the buffer is not replenished fast enough), so the buffer size affects how long the peak bit rate can be exceeded. Alternatively, an encoder sets bit allocations in other encoding scenarios.

FIG. 14 shows a technique (1400) for setting relative bit allocations for textured video content, dark smooth video content and other smooth video content. In particular, FIG. 14 shows a multi-pass rate control process in which an encoder sets bit allocations for smooth content relative to dark smooth content. A tool such as the encoder (400) explained with reference to FIG. 4 or other tool performs the technique (1400) as part of encoding.

First, the encoder initializes (1410) bit allocations for textured video content versus dark video content versus other smooth video content. In the example implementation, the encoder initializes QP, TACL, TACC, TACLD, and TACCD parameters to have initial values. TACL, TACC, TACLD, and TACCD can all have different initial values, or they can have the same initial value, depending on how the encoder is implemented. To increase the encoding quality of dark smooth content relative to other smooth content, the encoder can increase TACLD and TACCD relative to TACL and TACC, respectively.

TACLD=TACL+x (6), and
TACCD=TACC+x (7),

where x is an encoder setting, possibly exposed to the user, such as 1, 2, 3, etc. Another common approach is to have the luma control parameters vary relative to the chroma control parameters.

TACC=TACL/y (8),

where y is an encoder setting, possibly exposed to the user, such as 1.5, 2, 2.5, etc. Alternatively, the encoder establishes other relations between the control parameters. Although the encoder separately tracks TACL, TACC, TACLD, and TACCD, they can vary together. In one possible configuration, the encoder sets QP=4 and TACL=TACC=TACLD=TACCD=4 for the first pass encoding. Alternatively, the encoder initializes other and/or different control parameters.

Then, the encoder encodes (1420) the video using the initialized bit allocations. In the example implementation, the encoder encodes the video and keeps the values of the parameters QP, TACL, TACC, TACLD, and TACCD constant throughout first pass encoding. During the encoding, the encoder classifies units of the video as textured, dark smooth or smooth, and the encoder tracks the bits used encoding textured content, dark smooth content, and other smooth content.

The encoder adjusts (1430) bit allocations for the textured video content versus dark video content versus other smooth video content. In the example implementation, the encoder adjusts values of the QP, TACL, TACC, TACLD, and TACCD parameters depending on the results of the first pass of encoding. The encoder can adjust the control parameters in view of the actual bit counts from the first pass encoding and the target bit rate. In the example implementation, the encoder compares the total actual size A of the first pass encoding results to the target size T for the encoding, which is obtained by multiplying the target (average) bit rate by the time duration of the video sequence. The encoder computes the ratio of T to A: R=T|A. The encoder uses the ratio R to scale bit allocations for textured, dark smooth, and other smooth content for second pass encoding.

For example, if the target size T is 1 GB and the actual size A after the first pass is 2 GB, then the ratio R=0.5. The encoder uses the ratio R to scale the control factors QP, TACL, TACC, TACLD and TACCD. The encoder adjusts the control parameter QP under the assumption that texture bits scale inversely with QP. So, for this example, the encoder adjusts QP to be 4/0.5=8. For each frame, the desired average number of texture bits allocated in the frame is half of the number that was produced in the frame for the first pass. The encoder adjusts the control parameters TACL, TACC, TACLD and TACCD under the assumption that smooth bits scale directly with the target number of coefficients. So, for this example, for each frame, the encoder adjusts the number of smooth and dark-smooth bits allocated to the frame to be half what was produced in the frame for the first pass. To achieve this allocation, the encoder directly scales the control parameters TACL, TACC, TACLD and TACCD by the ratio R, multiplying the initial value of each of the TAC parameters by 0.5. If the initial values of TACL, TACC, TACLD, and TACCD are TACL=4, TACC=2, TACLD=6, and TACCD=4, the adjusted values are TACL=2, TACC=1, TACLD=3, and TACCD=2. Alternatively, the encoder adjusts the control parameters using another approach, for example, changing control parameters on a picture-by-picture basis or some other basis.

The encoder then encodes (1440) the video using the adjusted bit allocations. In the example implementation, the encoder uses the adjusted control parameters QP, TACL, TACC, TACLD and TACCD to control macroblock-level QP values for non-textured macroblocks and to control DZ sizes for blocks of dark smooth or other smooth content. The encoder uses the same encoding tools for the first pass and second pass (e.g., same transform, same texture metrics, same thresholds for textured/non-textured content) and can reuse information such as classifications. Alternatively, the encoder uses different encoding tools.

Alternatively, the encoder uses the adjusted values of the QP, TACL, TACC, TACLD, and TACCD parameters as guidelines in the second pass encoding, but varies QP as needed to meet target bit rate constraints or other encoding constraints during the second pass encoding. If the encoding constraints are not satisfied using the adjusted values of the QP, TACL, TACC, TACLD, and TACCD parameters, the encoder incrementally increases or decreases the QP values used at picture level and/or macroblock level for rate control.

After the second pass encoding, the control parameters TACL, TACC, TACLD, and TACCD can be fine-tuned, if desired, to control bit allocations on a scene-by-scene, frame-by-frame or other basis. The encoder then encodes the video using the fine-tuned control parameters.

In the example implementation, the encoder effectively controls bit allocation of textured, dark smooth and other smooth content, and can expose control of the bit allocation to a user. The encoder controls bits allocated to textured blocks by adjusting picture-level QP values, controls bits allocated to dark smooth blocks by adjusting TACLD and TACCD, and controls bits allocated to smooth blocks by adjusting TACL and TACC. Roughly speaking, encoding that uses a QP value of QP₁produces Z times as many bits as encoding that uses a QP value of QP₁*Z. (So, encoding using a QP value of 4 is expected to produces twice as many bits as encoding using a QP value of 8.) Likewise, encoding that uses a TAC value of TAC₁*Z is expected to produce roughly Z times as many bits as encoding that uses a TAC value of TAC₁. (So, encoding using a TAC value of 4 is expected to produce twice as many bits as encoding using a TAC value of 2.) Given these basic relations, the encoder modulates bit allocation for textured, smooth and dark smooth regions by adjusting the values of QP, TACL, TACC, TACLD and TACCD.

Alternatively, an encoder uses another approach to setting control parameters for adjusting encoding quality of textured content, dark smooth content, and other smooth content using a multi-pass approach.

VI. Alternatives

Although much of the preceding description focuses on block-based video encoding applications, the described techniques and tools have uses in areas other than video encoding. For example, the described techniques and tools for content classification can be used for image encoding and other types of image processing.

Having described and illustrated the principles of our invention with reference to various embodiments, it will be recognized that the various embodiments can be modified in arrangement and detail without departing from such principles. It should be understood that the programs, processes, or methods described herein are not related or limited to any particular type of computing environment, unless indicated otherwise. Various types of general purpose or specialized computing environments may be used with or perform operations in accordance with the teachings described herein. Elements of embodiments shown in software may be implemented in hardware and vice versa.

In view of the many possible embodiments to which the principles of our invention may be applied, we claim as our invention all such embodiments as may come within the scope and spirit of the following claims and equivalents thereto.

Claims

1. A method comprising: for each of plural units of a picture: analyzing content of the unit; andusing a computer, assigning a classification to the unit based at least in part on results of the analyzing the content of the unit;using the classifications to process the plural units, wherein the classification for at least one of the plural units is a first classification type, and wherein the classification for at least one of the plural units is a second classification type indicating darker content than the first classification type; andoutputting a result of the processing;wherein the assigning the classification to the unit includes:comparing energy of AC coefficients for the unit to a threshold, wherein the unit is classified as a third classification type if the energy is higher than the threshold, the third classification type indicating textured content, and the first and second classification types indicating smooth content.
2. The method of claim 1 wherein the assigning includes classifying the unit as the first classification type or the second classification type depending on average sample intensity value for the unit.
3. The method of claim 1 further comprising setting the threshold using a level of texture classification specified according to user input received through a user interface control.
4. The method of claim 3 wherein the setting uses a non-linear mapping of possible values of level of texture classification to possible values of the threshold.
5. The method of claim 1 further comprising setting one or more parameters to control bit allocation for the second classification type relative to the first classification type during encoding of the plural units.
6. The method of claim 5 wherein the one or more parameters include a target AC coefficient count for the encoding of the at least one of the plural units classified as the second classification type.
7. The method of claim 5 wherein the one or more parameters include separate control parameters for luma units of the second classification type and chroma units of the second classification type.
8. The method of claim 5 wherein the encoding includes: using the one or more parameters and the classifications, determining how to quantize the plural units; andquantizing the plural units.
9. The method of claim 5 wherein the setting the one or more parameters comprises: in a preliminary pass, encoding the plural units using initial values for the one or more parameters; andadjusting values of the one or more parameters based at least in part upon results of the encoding using the initial values in the preliminary pass.
10. A method of encoding plural video units, the method comprising: setting one or more parameters to control bit allocation for a first classification type relative to a second classification type during encoding of plural video units, the second classification type indicating darker content than the first classification type, wherein the setting includes: in a first pass, encoding, using a computer, the plural video units using initial values for the one or more parameters; andadjusting values of the one or more parameters based at least in part upon results of the encoding using the initial values in the first pass;in a second pass, encoding the plural video units using the adjusted values; andoutputting the encoded plural video units from the second pass in a bit stream.
11. The method of claim 10 wherein the one or more parameters include a first target AC coefficient count for the first classification type and a second target AC coefficient count for the second classification type.
12. The method of claim 10 wherein the one or more parameters include a first target AC coefficient count for luma sample values of the second classification type and a second target AC coefficient count for chroma sample values the second classification type.
13. The method of claim 10 wherein the adjusting the values of the one or more parameters includes changing the initial values based at least in part upon target bit rate and resulting bit rate of the encoding in the first pass.
14. The method of claim 10 wherein the first classification type and the second classification type indicate non-textured content, wherein the encoding in the first pass and the encoding in the second pass include classifying the plural video units, and wherein: the classification for at least one of the plural video units is a third classification type indicating textured content,the classification for at least one of the plural video units is the first classification type, andthe classification for at least one of the plural video units is the second classification type.
15. The method of claim 14 wherein the classifying uses a texture threshold, the method further comprising: receiving input that specifies a level of texture classification; andassigning a value to the texture threshold based at least in part upon the level, wherein the assigning uses a non-linear mapping of possible values of the level of texture classification to possible values of the texture threshold.
16. A video encoding system comprising: a user interface module adapted to receive user input that indicates a level of texture classification;an encoding control module adapted to set a texture threshold based at least in part upon the level of texture classification and using a non-linear mapping of possible values of level of texture classification to the possible values of the texture threshold, wherein the texture threshold is an AC energy threshold;one or more video encoding modules adapted to encode video using the texture threshold.
17. The system of claim 16 wherein the non-linear mapping embodies finer gradations for simpler content and coarser gradations for more complex content.
18. The system of claim 16 wherein the video encoding includes: classifying plural units of the video as textured, dark smooth or smooth; andusing the classifications to allocate bits between the plural units during the encoding.
19. The system of claim 16 wherein the one or more video encoding modules include a module adapted to set parameters to control bit allocation between textured content, dark smooth content, and other smooth content during the encoding.

US Referenced Citations (325)

Number	Name	Date	Kind
762026	Connstein	Jun 1904	A
4583114	Catros	Apr 1986	A
4679079	Catros et al.	Jul 1987	A
4774574	Daly et al.	Sep 1988	A
4821119	Gharavi	Apr 1989	A
4862264	Wells et al.	Aug 1989	A
4965830	Barham et al.	Oct 1990	A
4992889	Yamagami et al.	Feb 1991	A
5072295	Murakami et al.	Dec 1991	A
5128758	Azadegan et al.	Jul 1992	A
5136377	Johnston et al.	Aug 1992	A
5144426	Tanaka et al.	Sep 1992	A
5179442	Azadegan et al.	Jan 1993	A
5237410	Inoue	Aug 1993	A
5241395	Chen	Aug 1993	A
5253058	Gharavi	Oct 1993	A
5263088	Hazu et al.	Nov 1993	A
5301242	Gonzales et al.	Apr 1994	A
5303058	Fukuda et al.	Apr 1994	A
5317396	Fujinami	May 1994	A
5317672	Crossman et al.	May 1994	A
5333212	Ligtenberg	Jul 1994	A
5351310	Califano et al.	Sep 1994	A
5374958	Yanagihara	Dec 1994	A
5412429	Glover	May 1995	A
5452104	Lee	Sep 1995	A
5461421	Moon	Oct 1995	A
5473377	Kim	Dec 1995	A
5481553	Suzuki et al.	Jan 1996	A
5506916	Nishihara et al.	Apr 1996	A
5510785	Segawa et al.	Apr 1996	A
5537440	Eyuboglu et al.	Jul 1996	A
5537493	Wilkinson	Jul 1996	A
5539469	Jung	Jul 1996	A
5559557	Kato	Sep 1996	A
5565920	Lee et al.	Oct 1996	A
5587708	Chiu	Dec 1996	A
5606371	Gunnewiek et al.	Feb 1997	A
5623424	Azadegan et al.	Apr 1997	A
5631644	Katata et al.	May 1997	A
5654760	Ohtsuki	Aug 1997	A
5657087	Jeong et al.	Aug 1997	A
5663763	Yagasaki et al.	Sep 1997	A
5724097	Hibi et al.	Mar 1998	A
5724456	Boyack et al.	Mar 1998	A
5731836	Lee	Mar 1998	A
5731837	Hurst, Jr.	Mar 1998	A
5739861	Music	Apr 1998	A
5751358	Suzuki et al.	May 1998	A
5751379	Markandey et al.	May 1998	A
5761088	Hulyalkar et al.	Jun 1998	A
5764803	Jacquin et al.	Jun 1998	A
5781788	Woo et al.	Jul 1998	A
5786856	Hall et al.	Jul 1998	A
5802213	Gardos	Sep 1998	A
5809178	Anderson et al.	Sep 1998	A
5815097	Schwartz et al.	Sep 1998	A
5819035	Devaney et al.	Oct 1998	A
5825310	Tsutsui	Oct 1998	A
5835145	Ouyang et al.	Nov 1998	A
5835237	Ebrahimi	Nov 1998	A
5844613	Chaddha	Dec 1998	A
5850482	Meany et al.	Dec 1998	A
5867167	Deering	Feb 1999	A
5870435	Choi et al.	Feb 1999	A
5877813	Lee et al.	Mar 1999	A
5878166	Legall	Mar 1999	A
5880775	Ross	Mar 1999	A
5883672	Suzuki et al.	Mar 1999	A
5926791	Ogata et al.	Jul 1999	A
5969764	Sun et al.	Oct 1999	A
5970173	Lee et al.	Oct 1999	A
5990957	Ryoo	Nov 1999	A
6044115	Horiike et al.	Mar 2000	A
6049630	Wang et al.	Apr 2000	A
6058362	Malvar	May 2000	A
6072831	Chen	Jun 2000	A
6084636	Sugahara et al.	Jul 2000	A
6088392	Rosenberg	Jul 2000	A
6091777	Guetz et al.	Jul 2000	A
6104751	Artieri	Aug 2000	A
6118817	Wang	Sep 2000	A
6118903	Liu	Sep 2000	A
6125140	Wilkinson	Sep 2000	A
6148107	Ducloux et al.	Nov 2000	A
6148109	Boon et al.	Nov 2000	A
6160846	Chiang et al.	Dec 2000	A
6167091	Okada et al.	Dec 2000	A
6182034	Malvar	Jan 2001	B1
6212232	Reed et al.	Apr 2001	B1
6215905	Lee et al.	Apr 2001	B1
6223162	Chen et al.	Apr 2001	B1
6240135	Kim	May 2001	B1
6240380	Malvar	May 2001	B1
6243497	Chiang et al.	Jun 2001	B1
6249614	Kolesnik et al.	Jun 2001	B1
6256422	Mitchell et al.	Jul 2001	B1
6256423	Krishnamurthy et al.	Jul 2001	B1
6263022	Chen et al.	Jul 2001	B1
6263024	Matsumoto	Jul 2001	B1
6275614	Krishnamurthy et al.	Aug 2001	B1
6278735	Mohsenian	Aug 2001	B1
6292588	Shen et al.	Sep 2001	B1
6314208	Konstantinides et al.	Nov 2001	B1
6337881	Chaddha	Jan 2002	B1
6347116	Haskell et al.	Feb 2002	B1
6348945	Hayakawa	Feb 2002	B1
6356709	Abe et al.	Mar 2002	B1
6359928	Wang et al.	Mar 2002	B1
6360017	Chiu et al.	Mar 2002	B1
6370502	Wu et al.	Apr 2002	B1
6373894	Florencio et al.	Apr 2002	B1
6385343	Kuroda et al.	May 2002	B1
6389171	Washington	May 2002	B1
6393155	Bright et al.	May 2002	B1
6408026	Tao	Jun 2002	B1
6418166	Wu et al.	Jul 2002	B1
6438167	Shimizu et al.	Aug 2002	B1
6456744	Lafe	Sep 2002	B1
6463100	Cho et al.	Oct 2002	B1
6466620	Lee	Oct 2002	B1
6473534	Merhav et al.	Oct 2002	B1
6490319	Yang	Dec 2002	B1
6493385	Sekiguchi et al.	Dec 2002	B1
6519284	Pesquet-Popescu et al.	Feb 2003	B1
6526096	Lainema et al.	Feb 2003	B2
6546049	Lee	Apr 2003	B1
6571019	Kim et al.	May 2003	B1
6593925	Hakura et al.	Jul 2003	B1
6600836	Thyagarajan et al.	Jul 2003	B1
6647152	Willis et al.	Nov 2003	B2
6654417	Hui	Nov 2003	B1
6678422	Sharma et al.	Jan 2004	B1
6687294	Yan et al.	Feb 2004	B2
6704718	Burges et al.	Mar 2004	B2
6721359	Bist et al.	Apr 2004	B1
6728317	Demos	Apr 2004	B1
6731811	Rose	May 2004	B1
6738423	Lainema et al.	May 2004	B1
6747660	Olano et al.	Jun 2004	B1
6759999	Doyen	Jul 2004	B1
6760482	Taubman	Jul 2004	B1
6765962	Lee et al.	Jul 2004	B1
6771830	Goldstein et al.	Aug 2004	B2
6785331	Jozawa et al.	Aug 2004	B1
6788740	van der Schaar et al.	Sep 2004	B1
6792157	Koshi et al.	Sep 2004	B1
6795584	Karczewicz et al.	Sep 2004	B2
6801572	Yamada et al.	Oct 2004	B2
6807317	Mathew et al.	Oct 2004	B2
6810083	Chen et al.	Oct 2004	B2
6831947	Ribas Corbera	Dec 2004	B2
6862320	Isu et al.	Mar 2005	B1
6865291	Zador	Mar 2005	B1
6873654	Rackett	Mar 2005	B1
6876703	Ismaeil et al.	Apr 2005	B2
6882753	Chen et al.	Apr 2005	B2
6907142	Kalevo et al.	Jun 2005	B2
6909745	Puri et al.	Jun 2005	B1
6947045	Ostermann et al.	Sep 2005	B1
6975680	Demos	Dec 2005	B2
6977659	Dumitras et al.	Dec 2005	B2
6983018	Lin et al.	Jan 2006	B1
6990242	Malvar	Jan 2006	B2
7016546	Fukuhara et al.	Mar 2006	B2
7020204	Auvray et al.	Mar 2006	B2
7027506	Lee et al.	Apr 2006	B2
7027507	Wu	Apr 2006	B2
7035473	Zeng et al.	Apr 2006	B1
7042941	Laksono et al.	May 2006	B1
7058127	Lu et al.	Jun 2006	B2
7099389	Yu et al.	Aug 2006	B1
7099515	Lin et al.	Aug 2006	B2
7110455	Wu et al.	Sep 2006	B2
7162096	Horowitz	Jan 2007	B1
7200277	Joshi et al.	Apr 2007	B2
7289154	Gindele	Oct 2007	B2
7295609	Sato et al.	Nov 2007	B2
7301999	Filippini et al.	Nov 2007	B2
7307639	Dumitras et al.	Dec 2007	B1
7356085	Gavrilescu et al.	Apr 2008	B2
7463780	Fukuhara et al.	Dec 2008	B2
7471830	Lim et al.	Dec 2008	B2
7580584	Holcomb et al.	Aug 2009	B2
7738554	Lin et al.	Jun 2010	B2
7778476	Alvarez et al.	Aug 2010	B2
7801383	Sullivan	Sep 2010	B2
7869517	Ghanbari	Jan 2011	B2
7889790	Sun	Feb 2011	B2
7995649	Zuo et al.	Aug 2011	B2
20010048718	Bruls et al.	Dec 2001	A1
20020021756	Jayant et al.	Feb 2002	A1
20020044602	Ohki	Apr 2002	A1
20020118748	Inomata et al.	Aug 2002	A1
20020118884	Cho et al.	Aug 2002	A1
20020136308	Le Maguet et al.	Sep 2002	A1
20020154693	Demos et al.	Oct 2002	A1
20020186890	Lee et al.	Dec 2002	A1
20030021482	Lan et al.	Jan 2003	A1
20030053702	Hu	Mar 2003	A1
20030095599	Lee et al.	May 2003	A1
20030103677	Tastl et al.	Jun 2003	A1
20030108100	Sekiguchi et al.	Jun 2003	A1
20030113026	Srinivasan et al.	Jun 2003	A1
20030128754	Akimoto et al.	Jul 2003	A1
20030128756	Oktem	Jul 2003	A1
20030138150	Srinivasan	Jul 2003	A1
20030185420	Sefcik et al.	Oct 2003	A1
20030194010	Mukerjee et al.	Oct 2003	A1
20030206582	Srinivasan et al.	Nov 2003	A1
20030215011	Wang et al.	Nov 2003	A1
20030219073	Lee et al.	Nov 2003	A1
20030223493	Ye et al.	Dec 2003	A1
20040008901	Avinash	Jan 2004	A1
20040022316	Ueda et al.	Feb 2004	A1
20040036692	Alcorn et al.	Feb 2004	A1
20040090397	Doyen et al.	May 2004	A1
20040091168	Jones et al.	May 2004	A1
20040151243	Bhaskaran et al.	Aug 2004	A1
20040158719	Lee et al.	Aug 2004	A1
20040190610	Song et al.	Sep 2004	A1
20040202376	Schwartz et al.	Oct 2004	A1
20040228406	Song	Nov 2004	A1
20040264568	Florencio	Dec 2004	A1
20040264580	Chiang Wei Yin et al.	Dec 2004	A1
20050002575	Joshi et al.	Jan 2005	A1
20050008075	Chang	Jan 2005	A1
20050013365	Mukerjee et al.	Jan 2005	A1
20050013497	Hsu et al.	Jan 2005	A1
20050013498	Srinivasan et al.	Jan 2005	A1
20050013500	Lee et al.	Jan 2005	A1
20050015246	Thumpudi et al.	Jan 2005	A1
20050015259	Thumpudi et al.	Jan 2005	A1
20050024487	Chen	Feb 2005	A1
20050031034	Kamaci et al.	Feb 2005	A1
20050036698	Beom	Feb 2005	A1
20050036699	Holcomb et al.	Feb 2005	A1
20050041738	Lin et al.	Feb 2005	A1
20050052294	Liang et al.	Mar 2005	A1
20050053151	Lin et al.	Mar 2005	A1
20050053158	Regunathan et al.	Mar 2005	A1
20050084009	Furukawa et al.	Apr 2005	A1
20050084013	Wang et al.	Apr 2005	A1
20050094731	Xu et al.	May 2005	A1
20050105612	Sung et al.	May 2005	A1
20050105622	Gokhale	May 2005	A1
20050123274	Crinon et al.	Jun 2005	A1
20050135484	Lee et al.	Jun 2005	A1
20050147163	Li et al.	Jul 2005	A1
20050152451	Byun	Jul 2005	A1
20050180500	Chiang et al.	Aug 2005	A1
20050180502	Puri	Aug 2005	A1
20050190836	Lu et al.	Sep 2005	A1
20050207492	Pao	Sep 2005	A1
20050232501	Mukerjee	Oct 2005	A1
20050238096	Holcomb et al.	Oct 2005	A1
20050254719	Sullivan	Nov 2005	A1
20050259729	Sun	Nov 2005	A1
20050276493	Xin et al.	Dec 2005	A1
20060013307	Olivier et al.	Jan 2006	A1
20060013309	Ha et al.	Jan 2006	A1
20060018552	Malayath et al.	Jan 2006	A1
20060034368	Klivington	Feb 2006	A1
20060038826	Daly	Feb 2006	A1
20060056508	Lafon et al.	Mar 2006	A1
20060071825	Demos	Apr 2006	A1
20060083308	Schwarz et al.	Apr 2006	A1
20060088098	Vehvilainen	Apr 2006	A1
20060098733	Matsumura et al.	May 2006	A1
20060104350	Liu	May 2006	A1
20060104527	Koto et al.	May 2006	A1
20060126724	Cote et al.	Jun 2006	A1
20060126728	Yu et al.	Jun 2006	A1
20060133478	Wen	Jun 2006	A1
20060133479	Chen et al.	Jun 2006	A1
20060140267	He et al.	Jun 2006	A1
20060165176	Raveendran et al.	Jul 2006	A1
20060188014	Civanlar et al.	Aug 2006	A1
20060197777	Cha et al.	Sep 2006	A1
20060227868	Chen et al.	Oct 2006	A1
20060238444	Wang et al.	Oct 2006	A1
20060239576	Mukherjee	Oct 2006	A1
20060245506	Lin et al.	Nov 2006	A1
20060256851	Wang et al.	Nov 2006	A1
20060256867	Turaga et al.	Nov 2006	A1
20060257037	Samadani	Nov 2006	A1
20060268990	Lin et al.	Nov 2006	A1
20060268991	Segall et al.	Nov 2006	A1
20070002946	Bouton et al.	Jan 2007	A1
20070009039	Ryu	Jan 2007	A1
20070009042	Craig et al.	Jan 2007	A1
20070053603	Monro	Mar 2007	A1
20070081588	Raveendran et al.	Apr 2007	A1
20070140333	Chono et al.	Jun 2007	A1
20070147497	Bao et al.	Jun 2007	A1
20070160138	Wedi et al.	Jul 2007	A1
20070160151	Bolton et al.	Jul 2007	A1
20070189626	Tanizawa et al.	Aug 2007	A1
20070201553	Shindo	Aug 2007	A1
20070230565	Tourapis et al.	Oct 2007	A1
20070237221	Hsu et al.	Oct 2007	A1
20070237222	Xia et al.	Oct 2007	A1
20070237236	Chang et al.	Oct 2007	A1
20070237237	Chang et al.	Oct 2007	A1
20070248163	Zuo et al.	Oct 2007	A1
20070248164	Zuo et al.	Oct 2007	A1
20070258518	Tu et al.	Nov 2007	A1
20070258519	Srinivasan	Nov 2007	A1
20080008394	Segall	Jan 2008	A1
20080031346	Segall	Feb 2008	A1
20080068446	Barkley et al.	Mar 2008	A1
20080080615	Tourapis et al.	Apr 2008	A1
20080089410	Lu et al.	Apr 2008	A1
20080101465	Chono et al.	May 2008	A1
20080187042	Jasinschi	Aug 2008	A1
20080192822	Chang et al.	Aug 2008	A1
20080240250	Lin et al.	Oct 2008	A1
20080240257	Chang et al.	Oct 2008	A1
20080260278	Zuo et al.	Oct 2008	A1
20080304562	Chang et al.	Dec 2008	A1
20090207919	Yin et al.	Aug 2009	A1
20090213930	Ye et al.	Aug 2009	A1
20090290635	Kim et al.	Nov 2009	A1
20090296808	Regunathan et al.	Dec 2009	A1
20100177826	Bhaumik et al.	Jul 2010	A1

Foreign Referenced Citations (27)

Number	Date	Country
1327074	Feb 1994	CA
0932306	Jul 1999	EP
1465349	Oct 2004	EP
1871113	Dec 2007	EP
897363	May 1962	GB
05-227525	Sep 1993	JP
07-222145	Aug 1995	JP
07-250327	Sep 1995	JP
08-336139	Dec 1996	JP
10-336656	Dec 1998	JP
11-041610	Feb 1999	JP
2003061090	Feb 2003	JP
6-296275	Oct 2004	JP
2007-281949	Oct 2007	JP
132895	Oct 1998	KR
WO 9721302	Jun 1997	WO
WO 9948300	Sep 1999	WO
WO 0021207	Apr 2000	WO
WO 0072599	Nov 2000	WO
WO 0207438	Jan 2002	WO
WO 2004100554	Nov 2004	WO
WO 2004100556	Nov 2004	WO
WO 2005065030	Jul 2005	WO
WO 2006075895	Jul 2006	WO
WO 2006112620	Oct 2006	WO
WO 2007015047	Feb 2007	WO
WO 2007130580	Nov 2007	WO

Related Publications (1)

	Number	Date	Country
	20090245587 A1	Oct 2009	US

Classifying and controlling encoding quality for textured, dark smooth and smooth video content

Information

Patent Number

Date Filed

Date Issued

Inventors

Original Assignees

Examiners

Agents

CPC

US Classifications

Field of Search

US

International Classifications

Term Extension

Abstract

Description

Claims

US Referenced Citations (325)

Foreign Referenced Citations (27)

Related Publications (1)