The present invention relates to methods of writing composite 1-step holograms.
PCT applications WO01/45943 and WO01/42861 (Brotherton-Ratcliffe et al.) described a holographic printing system based on a rapid pulsed laser. This system was able to operate in two modes (“a dual-mode printer”). In the first mode, a holographic master hologram, also known as an H1, is produced. This hologram is then copied conventionally to a second and final hologram, known as the H2, by machines similar to those described by Grichine et al. (1997). In the second mode of operation, the final hologram is written directly.
In its first mode of operation, the digital data required by the system disclosed in WO01/45943 and WO01/42861 consists of ordinary perspective views that are easily generated by standard 3D commercial software packages. These images are then trivially distorted in order to compensate for inherent optical distortion that is present in the invention.
In its second mode of operation, the digital data required by the system disclosed in WO01/45943 and WO00/42861 are derived by applying various general mathematical transformations on the ensemble of the undistorted data-set that would be used for the generation of a hologram under the first mode of operation.
U.S. Pat. No. 5,793,503 (Haines et al.) described various transformations for the preparation of digital data for writing the hologram of a 3D computerized model using an ideal hologram printer. This arrangement concentrated on the treatment of data from a specialized 3D model but also treated the case of the generation of the required digital data from the conventional perspective views produced by standard commercial software.
The methods described in U.S. Pat. No. 5,793,503 are inappropriate for application to the type of holographic printers described in PCT applications WO01/45943 and WO01/42861. One reason for this is that such printers intrinsically possess very large optical distortion caused by a finite 5th order Seidel coefficient in the writing objective. This distortion, which generally varies from pixel to pixel in the case of non-static SLM printers should not be corrected for independently by sequential application of correction algorithms as this would lead to both image noise and computational disadvantage.
Another reason why the arrangement disclosed in U.S. Pat. No. 5,793,503 is inappropriate in the present context is that the 1-step holograms that are produced by the system disclosed in PCT applications WO01/45943 and WO01/42861 must, on display, generally be illuminated by a light from a point source which does not correspond geometrically to the recording illumination used within the printer and thus any proper method, for transforming perspective data into the required data, should be based on a general diffraction model. In its generality such a model must take into account such parameters as the emulsion and hologram substrate refractive indices as well as the recording and ray replay geometry.
Although the system disclosed in WO01/45943 and WO01/42861 represents a considerable advance over the prior-art it too suffers from various limitations. In particular, by not integrating the correction for optical distortion of the objective, into the data rearrangement transforms necessary for writing 1-step holograms, image quality is inevitably compromised. In addition, by only tracking the reference recording beam in one dimension, rather than two, the printer is fundamentally unable to produce large format holograms that are illuminated by close point-source lights. Finally, by seeking to correct for geometrical image distortions alone the prior-art printer suffers from increasing discoloration effects as a closer (and more realistic) point-source illumination is demanded.
According to an aspect of the present invention there is provided a method of writing a composite 1-step hologram, comprising:
According to another aspect of the present invention, there is provided a method of writing a composite 1-step hologram, comprising:
The digital data may be acquired from a real object and comprises a plurality of perspective views of the required hologram image. Alternatively, the digital data may be acquired from a virtual object and comprises a plurality of perspective views of the required hologram image.
The composite 1-step hologram may comprise a transmission hologram or a reflection hologram.
According to one embodiment, the single mathematical transformation generates a rectangular viewing window located in front of the hologram. The spatial light modulator may either be static whilst writing the hologram or the spatial light modulator may be moving whilst writing the hologram. The viewing window is preferably either of substantially similar size to the composite hologram or of different size to the composite hologram. The viewing window may be symmetrically located in front of the composite hologram or generally offset from the centre of the composite hologram. Preferably, either the viewing window is located at the same perpendicular distance from a given point in the holographic image as the camera plane is located from the corresponding point on the object from which the digital data is acquired or the viewing window is located at a certain perpendicular distance from a given point in the holographic image and the camera plane is located at a substantially different perpendicular distance from the corresponding point on the object from which the digital data is acquired.
The digital data may be generated by a real or virtual camera which generates either a plurality of apodized images which are centred in a frame which corresponds with the object which is to be reproduced by the hologram or a plurality of non-apodized images having frames which correspond with the object which is to be reproduced by the hologram, the frames being generally off-centred.
According to an embodiment the single mathematical transformation generates a scrolling viewing window located in front of the hologram.
According to another embodiment the single mathematical transformation generates a viewing window having a fixed size in the horizontal dimension and which scrolls in the vertical dimension, the viewing window being located in front of the hologram.
According to another embodiment the single mathematical transformation generates a viewing window having a fixed size in the vertical dimension and which scrolls in the horizontal dimension, the viewing window being located in front of the hologram.
Preferably either the camera plane is located at a certain distance from a point on the object and the viewing plane is located at substantially the same distance from a corresponding point in the holographic image or the camera plane is located at a certain distance from a point on the object and the viewing plane is located at substantially a different distance from a corresponding point in the holographic image.
According to one embodiment the composite hologram is formed using a variable angle reference recording beam. According to another embodiment the composite hologram is formed using a fixed angle or collimated reference recording beam. The composite 1-step hologram may be replayed using a point-source light or alternatively using collimated light.
Preferably, the single mathematical transformation additionally integrally corrects the digital data for the image distortion caused by the altitudinal and azimuthal reference beam angle(s) used to replay each holographic pixel of the hologram being different from the altitudinal and azimuthal reference beam angle(s) used to write each holographic pixel of the hologram.
Preferably, the single mathematical transformation additionally integrally pre-distorts the digital data so that the data written on to the spatial light modulator is distorted. The reference recording beam is preferably overcorrected using either an astigmatic or a non-astigmatic geometry.
Preferably, individual altitudinal and azimuthal reference recording angles are determined for at least a majority, preferably all, of the holographic pixels forming the hologram.
According to an embodiment the overlap of viewing windows of a plurality of holographic pixels is arranged to be maximised. The overlap of viewing windows of two diagonally opposed holographic pixels may be maximised. Preferably, either the angular resolution within the overall viewing window of the hologram averaged over the overall viewing window is maximised or the angular resolution at the periphery of the overall viewing window is maximised. The pre-distortion of the digital data and the step of overcorrecting the reference recording beam are preferably arranged such that chromatic discoloration is minimized.
According to an embodiment the single mathematical transformation additionally integrally corrects the digital data for the distortion caused by emulsion swelling of the substrate.
According to an embodiment the single mathematical transformation additionally integrally corrects the digital data for the distortion caused by the wavelength of light used to replay the hologram being different from the wavelength of light used to write the hologram.
According to a particularly preferred embodiment, a plurality of colour channels are provided. According to an embodiment a red and/or green and/or blue colour channel are provided. A spatial light modulator may be provided for each colour channel. The composite 1-step hologram preferably comprises a multiple colour hologram. The multiple colour hologram is formed using reference recording beams having a first geometry and the hologram is replayed with light rays having a geometry different to the first geometry.
The actual replay wavelength may be calculated as a function of altitudinal and azimuthal angles for at least a majority, preferably all, of the holographic pixels forming the hologram.
Linear chromatic coupling tensors for each colour channel may be calculated. A separate tensor μνTαβ is preferably calculated for each primary colour channel. Preferably, a corrected tensor μνTαβ is then calculated for each primary colour channel as a linear combination of the uncorrected component colour tensors μνTαβ each operated on by respective chromatic coupling tensors. For each holographic pixel each of the corrected tensors is preferably written to a separate spatial light modulator in such a way as to create a fully colour-corrected composite colour hologram.
For a double parallax case, the single mathematical transformation between the tensors kgIij and μνTαβ preferably consists of a reordering of the elements according to a set of single index laws of the form k=f1(α, β,μ, ν,P1, Q1, H1, λ), g=f2(α, β,μ, ν,P1, Q1, H1, λ), i=f3(α, β,μ, ν,P1, Q1, H1, λ) and is j=f4(α, β,μ, ν, P1, Q1, H1, λ), wherein the functions fn are general functions of the indicated indices, P1 are a set of parameters characterizing the physical characteristics of the hologram, Q1 are a set of parameters characterizing the optical properties of the hologram writing mechanism, H1 is a set of parameters characterizing the geometrical properties of the reference recording and reference replay beams and λ is the wavelength at which the hologram is recorded.
For a single parallax case, the single mathematical transformation between the tensors kIij, and μνTαβ preferably consists of a reordering of the elements according to a set of single index laws of the form k=f1(α, β, μ, ν, P1, Q1, H1, λ), i=f2(α, β, μ, ν, P1, Q1, H1, λ) and j=f3(α, β, μ, ν, P1, Q1, H1, λ), wherein the functions fn are general functions of the indicated indices, P1 are a set of parameters characterizing the physical characteristics of the hologram, Q1 are a set of parameters characterizing the optical properties of the hologram writing mechanism, H1 is a set of parameters characterizing the geometrical properties of the reference recording and reference replay beams and λ is the wavelength at which the hologram is recorded.
According to another aspect of the present invention there is provided a 1-step holographic printer, comprising:
According to another aspect of the present invention there is provided a 1-step holographic printer, comprising:
The preferred embodiment provides a class of methods that is capable of effectively and efficiently transforming 2D digital perspective views that have been derived from a standard 3D computer model into data which is then used by a holographic printer to produce a 1-step composite hologram. Unlike the arrangement disclosed in U.S. Pat. No. 5,793,503 which is primarily directed towards the generation of small holograms in 1-step printers, the preferred embodiment is directed primarily towards the generation of larger holograms using either 1-step or dual-mode printers.
According to the preferred embodiment, rainbow or reflection type holograms may be produced. Rainbow holograms of a single colour channel or many colour channels may be generated. Equally reflection holograms of one or more colours are provided for, in addition to achromatic holograms of both rainbow and transmission nature. Particular attention is paid to reflection holograms (both single and multiple-colour and both single and double parallax).
The SLM may be moved which is preferred for printers that print both 1-step and 2-step holograms (dual mode printers).
According to a preferred embodiment holograms may be written which have different replay and recording reference geometries.
According to another embodiment the image data is transformed and the recording reference beam within the printer is adjusted preferably two-dimensionally in order to optimize the final viewing window of the hologram and to produce a distortion-free image, given the intended display illumination.
According to another embodiment the image data is transformed in order to produce a distortion free image and the recording reference beam within the printer is adjusted preferably two-dimensionally in order to optimize the final viewing window of the hologram and to maximize angular image resolution, given the intended display illumination.
According to another embodiment the image data is transformed in order to produce a distortion free image and the recording reference beam within the printer is adjusted preferably two-dimensionally in order to optimize the final viewing window of the hologram, to maximize angular image resolution and/or to minimize chromatic discoloration, given the intended display illumination.
According to another embodiment only one single set of numerical index rules is formulated and applied to convert raw image data into data required by the printer SLM(s), the set integrating the basic data reordering required for 1-step holograms with correction for each of a variety of geometrically distorting effects.
According to another embodiment the intrinsic chromatic discoloration of a 1-step hologram, arising when the recording and replay geometries differ, is corrected for by numerical transformation of the image data.
Various embodiments of the present invention together with other arrangements given for illustrative purposes only will now be described, by way of example only, and with reference to the accompanying drawings in which:
a illustrates the recording geometry (single parallax) for a translating camera that is always pointing forwards and of fixed field of view (FOV);
b illustrates the hologram geometry (seen from top) for a fixed viewing window that is equal in shape and size to the actual hologram (simple translating camera, fixed SIM, single parallax);
a illustrates, for the case of a tracking SLM, the relationship between the horizontal position of the centre of the SLM relative to the centre of the objective and the horizontal position of the centre of the projected image of the SLM at the viewing plane, again relative to the centre of the objective;
a illustrates an offset geometry whereby the viewing window is offset from the centre of the hologram—case of a general rectangular window—seen from the top;
b illustrates an offset geometry whereby the viewing window is offset from the centre of the hologram—case of a general rectangular window & single horizontal parallax—seen from the side;
a illustrates the geometry for a point-source reference;
b illustrates the geometry for a point-source reconstruction;
c illustrates the geometry for the object point on recording;
d illustrates the geometry for the image point on reconstruction;
e illustrates the geometry for a point sink reference;
a illustrates the case of
b illustrates the case of
c illustrates the case of
WO01/45943 and WO01/42861 disclose a dual-mode holographic printer based on pulsed laser technology and WO01/29487 and WO02/29500 disclose a multiple-colour pulsed laser which may be employed in the holographic printer. The known holographic printer is capable of producing either final 1-step holograms or H1 master holograms for H2 transfer. Holograms can be either of the reflection or transmission type. They may have full or limited parallax. They may be monochrome, rainbow or full-colour. The printing speed of the printer is several orders of magnitude greater than other known holographic printers. In addition it is compact and hologram quality is independent of external environmental noise.
In order to understand the application of the present invention the salient features of the known holographic printer will now be reviewed. For simplicity and clarity the case of a single monochromatic laser will be considered.
After passing through the liquid crystal display, the beam traverses a linear polarizer that converts the SIM image from a polarization rotation image into amplitude modulation. Then the beam passes through the wide-angle objective 119 mounted on the motorized translation stage 120 with motor 163. This stage is used to control the position of the focused image of the SLM produced by the objective 119. The size of the minimum waist 166 of the object beam is controlled by the motorized stage 108 with motor 109. The object beam now comes to bear on the photosensitive material 162 here shown as film mounted on a roll/stage system. The motor 129 controls movement of the stage 123 towards and away from the position of minimum object beam waist. The rollers 124 and 125 control the horizontal movement of the film 162 in front of the object beam. The motor 128 controls the vertical movement of the film in front of the object beam. Motor 126 controls the motion of the rollers 124 and 125. Rollers 122 and 131 tension the film and control the horizontal angle that the film makes to the axial propagation vector of the object beam.
The reference beam is split from the main laser beam by the variable beamsplitter 101 controlled by motor 165. The beam 135 is directed to a mirror 136 whereupon it is reflected through an quasi-elliptical or rectangular aperture 137, an effective image of which is eventually created at the intersection of the reference beam with the holographic recording material, such quasi-elliptical or rectangular shape producing a defined circular or quasi-elliptical or rectangular reference footprint on the recording material as may be required by the type of hologram being written. The reference beam continues to the waveplate 138 which controls the polarization of the laser beam. The elements 139 and 141 with either 164 or 163 form a telescope that controls the size of the beam after 164/163 which is adjustable by the motorized stage 142 with motor 143. The beamsplitter switch 144 either directs the reference beam on the path 154 or onto the path 145. Path 145 is used to create transmission holograms whereas path 154 is used to create reflection holograms.
In the case of path 145 the reference beam passes through the lens 164 that produces an approximate image of the aperture 137 at the recording material surface. This lens also corrects for the slight divergence of the light produced by the lens 141. The divergence of the light after 164, which is ideally collimated, is thus controlled to within diffraction limits. Practically this means that for small reference beam size the beam will not be exactly collimated but that such departure from collimation will lead to an image blurring significantly less than that induced by the source size of the final hologram illumination source. Mirrors 146 and 149 now direct the reference beam onto its target to intersect the object beam at the surface of the holographic recording material. Motorized rotation stages 147 and 150 with motors 148 and 152 respectively and the linear translation stage 151 with motor 153 assure that different reference angles may be achieved for different placements and orientations of the recording material.
In the case of path 154 the reference beam passes through the lens 163 that produces an approximate image of the aperture 137 at the recording material surface. This lens also corrects for the slight divergence of the light produced by the lens 141. The divergence of the light after 163, which is ideally collimated, is thus controlled to within diffraction limits as above. Mirrors 155 and 156 now direct the reference beam onto its target to intersect the object beam at the surface of the holographic recording material, this time from the opposite side to the object beam. The motorized rotation stage with motor 159 and the linear translation stage 158 with motor 160 assure that different reference angles may be achieved for different placements and orientations of the recording material.
The known holographic printer can function in a variety of different modes.
In order to record an H1 transmission hologram perspective views of a real or computer generated object are loaded into the SIM one by one, a holographic pixel recorded, the recording material advanced and the process repeated for each image. For the case of the generation of a rainbow transmission master hologram a line of pixels is written onto the holographic recording material as illustrated in
Turning now to consider the preferred embodiment of the present invention, in one embodiment a computer is used to generate a three dimensional model of an object using a standard commercial computer program. Current computer programs can produce very lifelike models using a variety of sophisticated rendering processes that mimic real life effects. In addition advances in computer technology have now seen the calculation times, required for such programs to run, dramatically decreased. Three dimensional scanners using Moiré or other principles now permit the incorporation of real world 3-D images in such computer models. The storage memory required for such 3-D models is largely dependent on the texture maps used therein and hence computer files representing such 3-D models are usually relatively small and may be transmitted over the internet easily. In the preferred embodiment of the invention we use such 3-D computer models to generate a series of 2-D camera views from a virtual viewing plane as shown in
In another embodiment of the invention a real model is used instead of a computer representation and a real camera is employed to record individual photographs (either digitally or via photographic film that is subsequently digitized). In such a case
In both of the above cases restricted animation, which may be transferred to the final hologram, may be modeled by arranging that the model 900 moves in a defined sense (representing such animation) as different camera positions are selected on the plane 901, such camera positions following sequential monotonic trajectories on the plane. On observing the final hologram, an observer following such sequential monotonic trajectory in the observation space will perceive the animation.
Mathematical Definition of the 3-D Image Data
The preferred embodiment works by defining a set of 2-D views of a real or computer represented object 900 (
With reference to
where the integers k and g label perspective view locations on 901. A grid of (NK×NG) perspective views is thus envisaged to cover the plane 901 which has dimensions Ξ×Θ.
With reference to
A grid of (NI×NJ) pixels is thus envisaged to cover each perspective view with each such view having dimensions of QX×QY. It is useful to project the “perspective-view” plane (904 & 905 etc of
In the case that we wish to model a full-parallax 3-D scene we define the luminous intensity tensor kgIij. This tensor represents the totality of information describing the 3-D scene. It can either be provided by multiple photographic data or as the output of a standard commercial 3-D modeling program. In the case of horizontal parallax holograms the index g is fixed and we define the luminous intensity tensor as simply kIij.
Mathematical Definition of the Hologram
The hologram that will be written is composed of many individual holopixels (
where D represents the hologram (horizontal) width and R the hologram height.
Let us now imagine a viewing plane which is geometrically coincident with the camera plane, situated at a distance H from the hologram and laterally centred. In this case the Cartesian coordinates (ξ, ζ) that have been defined above may be used for both the treatment of the viewing plane and the camera plane.
In the case of horizontal parallax holograms, the view of the hologram is the same irrespective of the vertical position bar a simple image shift. In this case a line may be defined, which we will call the camera line, which is horizontal and which is vertically centred in the viewing plane. If an H1 hologram were being written, this line would coincide exactly with the physical strip-master. Our computer model would then supply us with NK discrete views at regular intervals along this line. The variable ζ would no longer be needed and the distance ξ from the LHS of the camera line to a given camera view would be given by equation 1.
We now assume that all optical distortion is absent from the printer objective. This enables us to establish the required formalism for the paraxial case. Later we will then generalize our mathematics to the non-paraxial finite-distortion case.
The parameters Π and Σ effectively define respectively the horizontal and vertical fields of view (FOV) of the printer writing head. The objective optics are almost always circularly symmetric but when combined with an SLM unit, we obtain different FOVs in the horizontal and vertical directions. The paraxial printer FOVs are related to the parameters Π and Σ by the following relations:
Here the subscript P refers to “Printer” and the subscripts “H” and “V” refer respectively to horizontal and vertical. The parameter H, as discussed above, is the distance between the hologram and the viewing/camera plane.
To recapitulate, in the case of a full parallax hologram our computer data will consist of NK×NG perspective views of the required hologram image. Generally we will express this information as NK×NG intensity matrices or by using the simple intensity tensor kgIij, which gives the intensity distribution, I, at the camera view pixel (i, j) for the horizontal camera position k and the vertical camera position g. In the case of a single (horizontal) parallax hologram our computer model will consist of kIij. Ultimately we wish to calculate the data that must be written to the SLM(s) which we refer to as the paraxial mask file intensity distributions and which we will denote in tensor form as μνSαβ.
Single Parallax Monochrome Reflection Hologram with Static SLM
There are several choices as to how physically a single parallax direct write hologram may be written. Firstly, we may decide to keep the SIM static or we may use an objective having a larger entrance pupil and opt to move the SIM within this pupil. Secondly various formats of computer data that effectively correspond to different (virtual) camera geometries may be used. A simple translating camera will produce a different intensity tensor kIij than a specially programmed camera for example. Finally, different viewing window geometries may be selected. If we elect to use the full FOV of the objective when writing each holopixel then we will have a different result than if we constrain our viewing window to be a well-defined rectangle.
All the above choices must be made according to the specific application at hand. In the following sections we will treat the most important major cases and derive for each geometry the pixel-swapping (or interpolatory) transformations necessary to convert the single parallax perspective tensor kIij into the mask tensor μνSαβ.
Non-Centred Image with Fixed Rectangular Viewing Window of Same Size as Hologram
In this section a computer model, kIij, will be assumed. This model has been derived by a simple translating camera which follows a horizontal trajectory through the mid-point of the hologram viewing plane (when one interchanges the hologram+viewing plane for the virtual object+camera plane—see FIG; 14a). The viewing zone of the hologram is now defined to be geometrically identical to the actual hologram and thus constitutes a rectangle having horizontal dimension D and vertical dimension R (
The viewing window is laterally displaced from the hologram by a distance H and hence the camera track is also laterally displaced from the virtual object by this same distance H. The (virtual) camera FOV (ΨCH) is now chosen such that the rectangular view of the camera at a distance H is of width 2D. Therefore, when positioned at the extreme LHS of the viewing line, the camera's view will just extend to the extreme RHS of the hologram and visa-versa as shown in
Since the hologram is single-parallax the vertical camera FOV is chosen as
and the printer FOV is chosen as in equation 8a. Now, at the holopixel defined by (X,Y) (see
from whence it is trivial to see that
Also from this equation and the relation ξ=X+w it follows that
which simplifies to
This equation states that the ray defined by the mask file index μ and emanating from the holopixel whose index coordinates are (α, β) intersects with the camera view defined by the index k as given in this expression. To understand which pixel in the perspective view file this ray corresponds to we use the relation X=D−w derived above from whence it is seen that
or more simply
Finally, from t he equation y=Y it follows that
or more simply
If we wish to limit the vertical FOV of the hologram as a function of the holopixel coordinate Y such as to create a viewing window of exactly the same dimensions as the actual hologram then we will have to impose the condition that
where, of course, Σ≧2R. This condition on v translates into the following condition on ν:
And thus the full pixel-swapping transformation may be written as follows:
In the case that no restriction is imposed on V then we have the case of a rolling vertical window with every holopixel having an identical vertical FOV given by
Note that this is the same as the paraxial (vertical) printer FOV of equation 8a
but it is generally different from the virtual (vertical) camera FOV=
In the case of no restriction on V we may write the pixel swap transformation more simply as:
These equations define how the digital computer model kIij consisting of conventional perspective views is converted into the mask file information μνSαβ that is necessary to write the holopixels one by one via the SIM.
We have implicitly assumed that all the parameters i, j, k, g, α, βμ and ν are integers. Generally this does not necessarily have to be the case and the above transforms are perfectly valid if we insist on a rational representation. Indeed the transformations given above generally require, in the calculation of μνSαβ with integer parameters α, β, μ and ν, a knowledge of kIij with rational values of k, i and j. Such a situation means that either a computer perspective model must be generated on a specialized mesh that leads to a uniform mesh once the data is transformed to mask space or multi-dimensional interpolation is used to calculate the values of kIij; with the required rational values of k, i and j. Both solutions are practical and the choice of which method to use depends on the perspective model software and the computer hardware. There is, however, a third solution to the above problem where all index parameters are integers. For instance, if we choose NM to be odd and we further require
then it follows that the transformations given above remain closed in the integer set. In such a case simplified mask transformations may be written. For the case of a well defined rectangular viewing window we may therefore write:
and in the case of no restriction on ν:
μνSαβ=kIij 0≦k≦NK=0 otherwise (35)
where
k=μ+α−NK, (36)
i=NM−μ+1, (37)
j=β (38)
Centred Image with Fixed Rectangular Viewing Window
In the above section the case of a computer perspective model which was generated by a simply translating camera has been treated. However, such a model is rather inefficient as 2× as much rendering data must be calculated as is actually required for the case of a hologram viewing window equal to the hologram size. For more realistic cases, as shall be shown further on, the factor is even greater than 2×. It is thus rather more logical to render only that part of the data that actually appears on the hologram.
This may be accomplished (see
where ΨCH is the horizontal camera field-of-view and ΨCV is the vertical camera field-of-view and we render only between ξ=0 and ξ=D. In this way the coordinate point x=0 of every perspective view coincides with X=0 and we may use QX=D instead of QX=2D as in the previous section. Thus, in
Proceeding as before and with reference this time to
Thus
from whence trivially it follows that
Also from this equation and the relation ξ=X+W it follows that
which simplifies to
This equation states that the ray defined by the mask file index μ and emanating from the holopixel whose index coordinates are (α, β) intersects with the camera view defined by the index k as given in this expression. To understand which pixel in the perspective view file this ray corresponds to, the relation x=X, derived above, is used—whence
or more simply
Finally, as before, from the equation y-Y it follows that
or more simply
Thus, for a proper rectangular viewing window the following transformation is derived:
In the case that no restriction is imposed on ν then we have the case of a rolling vertical window with every holopixel having an identical vertical FOV. In this case the pixel swap transformation may be written as:
As before certain constraints can be identified that restrict the above transformations to the integer set. Thus, for example, we may require NM to be odd as before and we may further require
In such a case a simplified mask transformation may be written where all indices are integer. For the case of a well defined rectangular viewing window it follows that:
and for the case of no restriction on ν
μνSαβ=kIij ∀ν when 0<k≦NK (63)
=0 otherwise
where
k=μ+α−NK, i=α, j=β (64-66)
Centred Image with Maximum FOV
In the previous two sections the horizontal size of the viewing window has been restricted to be the same physical dimension as the actual hologram. The advantage of doing this is that in the viewing zone you either see all the hologram or nothing. However, each holopixel is potentially capable of replaying a fixed FOV. If all of this FOV were to be used, the effect would be a scrolling horizontal window instead of an abrupt image termination (assuming of course that the observer is located at the viewing plane). This case may be treated by considering a camera track longer than previously considered.
Hence Ξ==D+T. We further set Π=T (defining the horizontal printer FOV). Finally, we set QX=D and QY=R (defining the centred camera). As in the section entitled “Centred Image with Fixed Rectangular Viewing Window” a centred camera (
where ΨCH is the horizontal camera field-of-view, ΨCV is the vertical camera field-of-view and rendering is performed only between X=0 and X=D.
As before and with reference this time to
Likewise the new equation ξ=T2X+X+w now leads to:
which simplifies to
The equations for i and j remain as in the section entitled “Centred Image with Fixed Rectangular Viewing Window” and hence the final mask transformation for maximum FOV may be written:
If we wanted to limit the vertical viewing window to be of a fixed height Γ whilst keeping a maximum horizontal FOV, we could also use (see
In passing it should be noted that this condition may equally well be applied to the transformations of equations 19, 31, 49 and 59 if our intention is to restrict the vertical window to a height Γ. One may want to do this, by example, to achieve a greater hologram brightness.
Rectangular Viewing Window of General Size
In the above sections the case of a rectangular window of the same size as the hologram and the case of a scrolling window of unrestrained field of view have been treated. In this section, for completeness, the case of an arbitrarily sized viewing window and an arbitrary recording FOV will be treated.
In our above discussions of rectangular windows certain key parameters, such as the hologram recording FOV and the distance from the viewing plane to the hologram, were chosen in such a manor as to use most efficiently the printer SLM. However, under real conditions it is possible that to restrain Π to be equal to exactly twice the size of the hologram, for instance, as we have discussed in our first two analyses, may at some times be inconvenient. If a rectangular viewing window of horizontal dimension L and vertical dimension Γ is considered then our previous discussions can be generalized to requiring only that Π≧D+L and that Σ≧R+Γ. A centred camera will be used as in the two previous sections above and so we will require that QX=D and QY═R. Since the horizontal dimension of the viewing window is L we shall require Ξ=L.
As before it should be noted that the recording or printer FOVs are defined by Π and Σ. The camera FOVs however are now somewhat different. We will require that
where, as before, ΨCH is the horizontal camera field-of-view, ΨCV is the vertical camera field-of-view and rendering is performed only between X═0 and X=D.
With reference to
The vertical window is dwelt with exactly as previously and hence the general mask transformation may be written as:
In passing we should note that usually one should try and arrange that Π=D+L in order to utilize the full horizontal resolution of the SLM.
Moving SIM, Centred Camera and General Rectangular Window
Up until now the case of a static SLM which remains fixed in the entrance pupil of the writing objective has been considered. However, when we wish to make a hologram with a well defined rectangular window, we may do so by moving the SLM within the entrance pupil of the objective. In this case, of course, the objective must have a larger entrance pupil relative to the SLM size in order to accommodate such motion. One of the principle advantages of this scheme is its use in dual-mode printers. Generally MW printers that write H1 type master holograms require a moving SIM (in order to use effectively the SIM resolution) and hence, by using this scheme in a dual function printer, exactly the same print-head assembly can be employed to write either a 1-step hologram or an H1 hologram master. This is not the case if we use a static SIM.
Let us assume as usual that our hologram is of horizontal size D and vertical size R and that our viewing window, displaced laterally from the hologram by a distance H, is of horizontal dimension L and vertical dimension r. If the maximum angular resolution possible is to be attained with our SLM, H must be chosen with reference to the FOV of the writing objective such that the virtual image of the SLM at the viewing plane is of horizontal dimension L.
Now one can choose to move the SLM only horizontally, or one can choose to move it in a two-dimensional fashion. If it is moved 2-dimensionally then all the vertical pixels of the SLM will be used only in the case that the vertical SIM image size in the viewing plane is Γ. If the vertical size is restricted such that the aspect ratio of the SLM no longer corresponds to the aspect ratio of the viewing window then generally we will only use a fixed percentage of the vertical SIM height and Σ≧Γ. In the case that we elect to only move the SIM horizontally and to permanently position it in the vertical centre of the objective pupil then either an unrestricted scrolling window can be used or ν can be restrained, as in the previous sections, so as to create a fixed rectangular viewing window.
Firstly it should be pointed out that for the case of a horizontal viewing window size of L, the writing objective must have an FOV of at least
In the unlikely event that the viewing window is taller than it is wider then we would require that
but it should be noted that this is rather unlikely in the context of a single parallax hologram. It should also be noted that these equations constitute absolute limits pertaining to the translation of the SLM within the objective pupil in either a strictly horizontal or a strictly vertical fashion. Slightly more severe criteria may be derived by examining the outer diagonal points of the SLM within the objective pupil.
Let us consider a 2-D displacement of the SIM by denoting the x and y coordinates of the centre of the projected image of the SLIM at the viewing plane by τ and η. We choose the origin of this coordinate system at the holopixel being written—namely(X,Y).
Note that the actual x and y coordinates ({haeck over (τ)} and {haeck over (η)}) of the centre of the SIM relative to the centre of the paraxial objective are just linearly proportional to τ and η (i.e. {haeck over (τ)}=aτ and {haeck over (η)}=aη where is the constant a is related to the characteristics (magnification) of the objective—see
From
Thus we may write the mask transformation for the case of a 2-D moving SLM, centred virtual camera and of general rectangular viewing area as
If we do not wish to limit the hologram vertical FOV and hence to have a vertical scrolling window, then we may decide to fix the SLM vertically in the objective mid-plane and move it only horizontally. The above transform will then also apply with the exception'that there is no restriction on V.
If we demand that NK=NM, NI=NA and NJ=NB we see that the rules for k, i and j become trivially k=μ, m i=α and j=β.
Full Parallax Monochrome Reflection Hologram with Static SLM
Transformations derived for the single parallax case may be generalized with the greatest of ease to the full parallax case.
Non-Centred Image with Fixed Rectangular Viewing Window of Same size as Hologram for the Full-Parallax Case
We should first point out that a non-centred virtual camera of fixed FOV and which only translates in two dimensions provides a poor solution to any real design. This is because the computer is forced to generate at least four times the perspective information than is actually required.
As we have discussed before,
Since the horizontal viewing width is D we must require that Ξ=D. Since the vertical viewing height is R we likewise require that Θ═R. Our choice of a translating virtual camera imposes the constraints QX=2D and QY=2R and finally we will choose H such that Π=2D. Note that Σ≧2R as generally the aspect ratios of SLM and viewing zone will not be the same.
b and the logic of the section entitled “Non-Centred Image with Fixed Rectangular Viewing Window of Same Size as Hologram” can now be used to derive transformation rules for the indices k and i. These rules are identical to those given in equations 12 and 14. Referring now to
in the vertical plane. These relations give us equations for g and j:
Thus the full-parallax translating camera mask transformation may be written as
In the case that Σ=2R and both NM and NV are odd and further
a much simplified transform may be written:
μνSαβ=kgIij when 0<g≦NG and 0<k≦NK=0 otherwise (111)
where
k=μ+α−NK, (112)
i=NM−μ+1, (113)
g=ν+β−NG, (114)
j=NV−ν+1. (115)
Centred Image with Fixed Rectangular Viewing Window of same size as Hologram for the Full Parallax Case
We now generalize the concept of the centred camera to two dimensions. We thus change the camera apertures such that only data obeying the constraints 0≦X≦D and 0≦Y≦R is rendered. As usual we will choose H such that Π=2D.
The equations governing k and i remain as in the section entitled “Centred Image with Fixed Rectangular Viewing Window”.
However, as in the previous section, since the aspect ratio'of the hologram viewing zone will not in general be the same as the aspect ratio of the SLM, the vertical view needs more care (see
We now set Θ=R=Qy. We see that the single-parallax centred-camera relations u=D+iw, ξ=X+w and x=X in the horizontal plane are now replaced with
ζ=Y+w and y=Y in the vertical plane. These relations give us the new equations for g and j:
The full-parallax centred-camera mask transformation may now be written as
In the case that Σ=2R and both NM and NV are odd and further
a much simplified transform may be written:
μνSαβ=kgIij when 0<g≦NG and 0<k≦NK=0 otherwise (125)
where
k=μ+α−NK (126)
i=μ (127)
g=ν+β−NG (128)
j=ν (129)
The Maximum FOV Case for Full-Parallax Holograms
We now generalize the section entitled “Centred Image with Maximum FOV” to the case of horizontal and vertical scrolling windows of maximum FOV. With reference to
and we choose to render only those points obeying the constraints 0≦X≦D and 0≦Y≦R. We choose H such that π=T (i.e. one chooses H so as to arrive at the required width of the viewing zone=D+T which of course depends on the horizontal printer
Other parameters are Ξ=D+T, Θ=R+, QX=D and QY=R. The equations for the horizontal plane are therefore as before.
The mask transformation for maximum FOV may therefore be written as:
As before, if we wanted to write a hologram with the maximum horizontal FOV but to limit the vertical window to a fixed height, Γ, then we could apply the conditional transform of equation 76.
Or, in other words:
This would use the same image data as was used in equations 132 to 136 but would simply convert the vertical scrolling window into a rectangular window.
An easier way to limit the vertical window to a fixed height, Γ would be to re-render and incorporate the image clipping within the g index swap rule. In this case Ξ=Γ which would be considerably more (computationally) efficient over the previous criteria that Θ═R+Σ. In this case the mask transform becomes
General Rectangular Viewing Zone for Full Parallax
Using exactly the same logic, it is trivially clear that the single parallax case of a general rectangular window may be generalized to full parallax. Assuming again a centred camera with QX=D and QY=R and a rendering window defined by Ξ=L and Θ=Γ, the mask transformation may then be written:
The Case of a Moving SIM for Full-Parallax Holograms In the light of the above sections it will be clear to a person skilled in the art how the single parallax analysis already presented for a moving SIM may be generalized to the full-parallax case.
Offset Window Geometry
In all of the above discussions the viewing window, whether of a fixed rectangular topology or whether of a scrolling nature has been contemplated to be centrally placed in front of the hologram. However, in a commercial printer that must print large holograms, it is likely that sub-sections of the final hologram will have to be printed individually and then assembled. Hence we might consider dividing a lmxlm hologram up into 4 pieces of 50 cm×50 cm each. Clearly a discussion of the printing of these sub-segments entails a discussion of a general viewing window being offset from the respective hologram segment.
Due to the importance of this topic two explicit examples will be presented—one for the single parallax case and one for the double parallax case. It will then be evident to someone skilled in the art how this technique may be generalized to the various other viewing and recording geometries hereto above presented.
The first example that will be discussed in detail relates to a single parallax sub-hologram of size D×R with a generalized rectangular viewing zone of size L×Γwhose centre is generally laterally offset from the centre of the hologram by ωx in the x direction and by ωy in the y direction.
where as usual data is only rendering between 0≦X≦D and 0≦Y≦R.
It is worth pointing out here that the render data for an entire composite hologram may well be generated at one time and then this data may be sorted into data that is relevant to the various respective sub-holograms. Alternatively, as we have envisaged in equation 143, data is rendered individually for each sub-hologram.
With reference to
By analysis of
or in terms of indices
The mask transformation for a single parallax hologram with an offset rectangular window may therefore be written as follows:
It should be noted that the horizontal printer FOV must always be larger or equal to its counterpart camera FOV in this case. The vertical printer FOV will also need to be significantly larger than its counterpart camera FOV (see
We may now trivially generalize the above example to the full parallax case. The mask transformation (note that we require Θ=Γ) now becomes:
Note that in this case we require
In addition both the horizontal and vertical printer FOVs should be larger or equal to their counterpart camera FOVs.
Viewing Plane Different From Camera Plane
Up until now it has been assumed that the viewing plane is collocated at the camera plane. However it is possible to generalize to the case where these two planes are not collocated. Using our intuition or more formally, Huygens' principle, we see that in fact a full-parallax hologram does not need to be viewed at the camera plane in order to be undistorted. In contrast, a single-parallax hologram will inevitably show distortion if the viewer does not view the hologram at the camera plane. Depending on the size and depth of the single parallax hologram in question such distortion may either be important or in some cases, negligible.
In the case where digital image data is already available at one camera distance, it may sometimes make sense to use such data for the creation of a hologram having a different viewing window position rather than re-rendering. In this case the teaching presented hereto above will allow someone skilled in the art to derive similar equations to those already presented but covering the more general case of non-collocated camera and viewing planes.
We may also consider the possibility of defining an astigmatic viewing window, whose boundaries focus at different distances in the vertical and horizontal. For example the following transformation would describe the case of a single parallax hologram with an astigmatic rectangular viewing window with the horizontal window focus being located on the camera plane at a distance H from the hologram and the vertical window focus being located at a distance Hv
This transformation might be used when we wished the vertical window to have light scrolling characteristics at the (horizontal) viewing plane with a consequently less severe vertical window collapse going away from and towards the hologram. Such windows are particularly useful for highly asymmetric holograms where the width, D is very different from the height, R.
Finally paraxial mask transformations may be derived for generalized viewing window topologies (circles—ellipses—multiple viewing zones) and windows possessing generalized scrolling characteristics. In addition curved surface viewing and camera planes and the case of cameras that do not point always in the same direction are easily treated using the same formalism as hereto described.
Objective Distortion
In any real-world optical write-head of large FOV, there is inevitably significant optical distortion. Usually the predominant cause of this is aberration of the objective associated with a finite 5th Seidel coefficient. If we compare normalized object and image planes of a given objective as in
and f is a single-valued 1-dimensional function that describes the distortion. These equations may also be interpreted as a transformation from real to paraxial object planes. By replacing the expressions
in all preceding equations we may thus derive mask-file transformations appropriate for use in write-heads with finite distortion. First let's take the case of the section entitled “Non-Centred Image with Fixed Rectangular Viewing Window of Same Size as Hologram” that treats single-parallax holograms. Here Π=2D QX and QY=R. Thus
The equation ξ=X+w then leads to
Similarly the equation x=D−w leads to
and the equation y=Y leads to
Likewise the vertical condition now becomes
We therefore see that equations 19-22 are replaced by the following equations that are now valid for the finite objective distortion case:
Note that for clarity S has now been replaced by T to distinguish data that has been corrected for optical distortion (T) over the data S which has not. Equation 24 remains invariant. Note also that equations 25-27 generalize to equations 157-159 and equations 31-38 are not relevant for the finite distortion case.
We may also generalize, for the finite distortion case, the (single parallax) equations presented in the section entitled “Centred Image with Fixed Rectangular Viewing Window”. Equation 49 changes to equation 156. Equation 50 changes to equation 157. Equations 51 and 52 remain invariant as does equation 53. Equation 54 changes to equation 157. Equations 55 and 56 remain invariant.
Likewise we may generalize the (single-parallax) case of the section entitled “Centred Image with Maximum FOV” for finite distortion.
Here we start with the modified k equation:
from which we see that
Equations 72, 74 and 75 thus remain invariant whereas equation 73 transforms to equation 161. We may derive the analogue of equation 76 by noting that now
from whence we see that
The other part of equation 76 follows in the same fashion whereupon we see that this equation transforms to
Equations 73-80 transform as equations 73-75.
As a further example of the generalization of the paraxial mask transformations to their finite distortion counterparts we note that equation 87 must be replaced by
Thus equation 85 then changes to equation 164. Equation 87 changes to equation 168 and equations 88 and 89 remain invariant.
The full-parallax offset-window transformation likewise generalizes to the following form:
Following the technique that we have described and illustrated above, it will be clear to those skilled in the art how all single and double parallax paraxial mask transformations may be generalized to their finite distortion counterparts for the case of a static SLM.
In the case that the SIM is moved during the writing of the hologram equations 144 to 146 are modified.
We note that, as before, f is a simple single-valued one dimensional function. We also note that τ is a function of α and that η is a function of β. We may thus define the tensor ραβμν as representing the distortion function at every point of interest.
As before, by replacing the expressions
in all preceding equations we may thus derive mask-file transformations appropriate for the case of a moving SLM and for use in write-heads with finite distortion. Note that τ and η are given by equations 92 and 93.
All static and moving SLM mask-file transformations presented hitherto may be converted in the above fashion to finite distortion mask-file transforms. In practice only these finite distortion transforms are of practical use. One might imagine that an alternative procedure would be to employ paraxial transforms and then to reapply other transforms to correct for finite optical distortion.
For instance, if we define {haeck over (μ)}{haeck over (ν)}Tαβ as representing the mask information corrected for finite distortion then the following transformation will convert the paraxial mask information S into the required corrected form T:
However, such sequential application of transforms is undesirable and would lead to a significant increase both in computational time and most importantly to interpolation error.
The reason that sequential transforms act to significantly increase interpolation error is that usually paraxial mask-file transforms must be computed using a truncation type of interpolation. This is because the ‘magic number’ representation presented earlier is rarely flexible enough for practical commercial use. Thus, in sequential application, the index rules for the integers i, j, k and g are truncated from a rational representation to an integer representation, giving rise to a truncation error. Then later the index rules of 168b and 168c exacerbate the error by converting a generally irrational RHS containing the prior truncation error into an integer RHS. Additionally the constraint that third-party 3D software programs must be used implies that specialized non-Cartesian meshes are not an option. Even if more advanced forms of interpolation are used in the calculation of S from I (over and above simple truncation methods) one inevitably incurs an interpolative error. By using sequential transforms such errors are compounded whereas by formulating a single finite distortion mask transformation only one interpolation error is produced. It is thus vital that a single transform is derived that treats both the ideal paraxial mask-file transform and the write-head optical distortion.
Generalization to Multiple Colour
All of the mask transforms above are valid for each and every colour channel employed in the holographic printer. Note that each colour channel will have its own distortion function and as such ρμν and ραβμν will be different for each colour.
Other Distortions
Many other image distortions may arise in a holographic printing system. Amongst these are distortions caused by emulsion swelling due to chemical processing, replay wavelength not equal to record wavelength, reference Beam angle different on replay to recording, object Beam axial angle different on replay to recording and further holographic transferring.
Many of these distortions are mathematically similar to the objective distortion discussed above (although they may lack that distortion's rotational symmetry properties and they may be very different for each colour channel). This is, because such distortions usually act only to change the index swapping laws in the mask transformation. Therefore such distortions may be incorporated into a single mask transformation just as we have discussed above for the case of objective distortion. By deriving a single finite distortion mask transformation that incorporates all distortions inherent in the printing system we benefit greatly in terms of interpolation noise and computational speed.
Numerical Pre-distortion of Image Data for Compensation of Disparity between Recording and Replay Geometries
It has been seen above how fundamental (perspective view) image data must be transformed in order to write 1-step holograms. In particular the importance of including, in the definition of a single transform, all of the printing system's distortions has been underlined.
The objective distortion that has been discussed at length above is critical to the operation of the class of printers under discussion (i.e. that are suitable for writing medium to large format holograms). It is thus impossible to “minimize” this distortion as the performance of the printer is directly related to its magnitude. However, this is not necessarily the case with the other subsidiary distortions that we have introduced. The most important of these in 1-step holographic printing devices is usually the disparity of the recording and replay reference ray geometries. This arises due to the fact that larger holograms are inevitably lit by point sources and yet the simplest and most elegant solution for the recording is a collimated reference beam.
One approach that has been advocated in PCT applications WO 00/29909 and WO 0029908 is to essentially eliminate all distortions caused by disparity in the recording and illumination reference ray geometries such that in fact no numerical compensation for this distortion is required. However this leads to a complex mechanical solution for the two dimensional control of the recording reference beam and, as we shall see below, to non-optimum viewing characteristics of the final hologram.
We disclose here two solutions to the present problem relevant to reflection type holograms. The first uses a simple fixed and collimated reference beam for recording. Mask transforms are then defined that take into account this fact and the required replay geometry (usually a point source at some distance). The second solution utilizes a combination of deliberate image pre-distortion (integrated into a mask transform as above) and deliberate over-correction of the recording reference ray. Whilst being electromechanically more complicated this second solution enjoys the merit of producing holograms with superior viewing characteristics.
Geometrical Ray Analysis of a Hologram In order to understand how the digital image data that are used to record a hologram may be pre-distorted such that the final holographic image appears undistorted when viewed under a chosen lighting geometry, we will now study exactly how the holographic image is distorted when recording and replay geometries are different.
Model
Firstly let us use a standard right-handed spherical coordinate system centred on a given holopixel as depicted in
k2[sinφisin θi+sinθcsinφc]ext=k1[sinθosinφo+sinθrsinφr]ext (175)
k2[cosφisinθi+sinθccosφc]ext=k1[sinθo cosφo+sinθrcosφr]ext (176)
k2[cosθi+cosθc]int=αk1[cosθo+cosθr]int (177)
nsinθint=sinθext (178)
Here k1 is the wavenumber at recording and k2 is the wavenumber at replay. The parameter α represents a factor describing how much the emulsion is swollen before recording. The subscript ‘ext’ refers to the θ angles just outside the emulsion layer whereas the subscript ‘int’ refers to θ angles inside the emulsion. Equation 178, which is Snell's law, connects these two types of angles. Note that in the geometry that we have chosen φ is invariant across the emulsion/air interface.
Equations 175-177 may be derived in a number of ways, the simplest being to demand that
where the quantities
are the respective wavevectors, {right arrow over (K)} is the normal fringe plane vector, a is the emulsion swelling matrix and all angles are internal. Equations 175-177 then correspond to the x, y and z components of this equation. Note that the x and y components are invariant under internal-external angle transformations. Note also that the z component is independent of azimuthal coordinate. This is because z is orthogonal to {circumflex over (φ)}.
Now let us superimpose a right-handed Cartesian system on our spherical system as depicted in FIGS. 29(a-d). Note that the origin (x,y,z)=(0,0,0) coincides with the centre of the 1-step hologram. The origin of the spherical system, however, coincides with the holopixel under scrutiny.
The plane (x,y,z=hr) is the recording plane. The plane (x,y,z=hv) is the viewing plane. The plane (x,y,z=0) is the hologram plane. The point (Xh,yh,0) is the holopixel of interest. The point (x1,Y1,hr) represents the intersection of the recording object ray (that intersects with the holopixel and the actual object point) with the recording plane. The point (x2,y2,hv) represents the intersection of the replayed image ray (emanating from the holopixel) with the viewing plane.
The point (Xc,yc,zc) is the location of the point source illumination on replay. Note that zc>0 and that yc is negative for the given form of equations 175 to 178. Initially we will treat the case of a collimated reference beam in which case we will just describe this set of rays by the spherical coordinate θr. However, later on we will examine the case where the reference beam on recording is changed during the process of writing. In this case the intersection of the various rays will form a point sink at (xr, yr, zr) with zr positive (and yr negative) (see
Collimated Reference Wave Geometry
Now let us write down the various relations that exist between representations in the spherical and Cartesian coordinate systems:
xh=xc−rcsinθcsinφc (179)
yh=yc+rcsinθccosφc (180)
xh=x1−r1sinθosinφo (181)
yh=y1+r1sinθocosφo (182)
xh=x2−r2sinθisinφi (183)
yh=y2+r2sinθicosφi (184)
We can trivially rearrange these expressions to give equations for the terms present in equations 175 to 176:
Now let us combine equations 177 and 178:
Expressions for sinθ0, sinθ1 and sinθc may be obtained from equations 179 to 184. Specifically
Hence substituting equations 192 to 194 into 191 we obtain
Note that θr has been left explicitly in this expression as the case of a collimated recording reference beam, characterized by zero azimuthal angle and a given altitudinal angle θr, is being discussed. We now introduce the following variables
ε=xh−x1 (196)
τ=xh−x2 (197)
γ=y2−yh (198)
σ=y1−yh (199)
μ=yc−yh (200)
ξ=xh−xc. (201)
Using these expressions in equations 185-190 we can rewrite equations 175 and 176:
Dividing these two equations we eliminate k1 and k2:
Now let us substitute equations 196 to 201 into 195:
Using equation 202 we can now eliminate k1 and k2:
We must now write down relations describing the radial co-ordinates r1, r2 and rc. From
rc2=ξ2+μ2+zc2 (207)
r12=ε2σ2+hr2 (208)
r22=τ2+γ2+hv2 (209)
The (γ, τ) Equations
Equations 204, 206, 207, 208 and 209 now constitute a complete set of equations for the variables γ and τ. These two variables respectively tell us the y and x co-ordinates of the intersection of the diffracted ray emanating from the holopixel at (xh, yh, 0) with the viewing plane, on replay of the hologram. The equations allow us to thus calculate γ and τ if we know the geometry of the playback light and all the recording geometry. The equations are quartic in nature and may best be written in parametric form as follows:
Here we have used R=r2 for simplicity and
As we have the above equations 44-46 are quartic and hence have 4 solutions. Two of these solutions have negative values of R and hence we will disregard these as we have adopted the convention of positive R in the definition of our spherical coordinate system in order to maintain its single-valuedness. The first solution for positive R is:
The second solution for positive R is:
The various Ω parameters referred to in the above equations are simple algebraic functions of a,b,c,d,g and n. We do not list these functions here for reasons of length and for the reason that their derivation will be evident for a person skilled in the art. For example we have derived these functions ourselves using a commercial symbolic manipulation program.
Even though there are two positive R solutions to the equations 210 to 212 only one is interesting. We must choose the root by requiring that k2 as defined by equation 203 be non-zero and positive valued.
In this section a set of algebraic equations has been derived that define the x and y replay ray intersections with the viewing plane of the hologram. These equations assume a certain hologram recording geometry (we have used the case of a collimated plane recording wave above) and a certain hologram replay geometry (generally non-collimated).
The (γ, τ) equations will now be applied in order to analyze exactly how the form of the hologram viewing window distorts due to the disparity between recording and replay geometries.
The hologram, in
Whilst the viewing distance is of the order of 1.5 times the largest dimension of the hologram and the illumination light is situated at a distance of around 3 times the largest dimension of the hologram it is possible to pre-distort the digital image data in order to obtain an undistorted hologram (when the hologram is illuminated by a point source). Beyond the above cited constraints it becomes impractical to use pre-distortion as the only method to make large undistorted holograms and some manipulation of the reference beam must take place (see
As we have previously discussed Klug et al. use exact reference beam tracking in order to circumvent the need to pre-distort the digital image data. This technique cites such a method as overcoming the prohibitive computational load of pre-distortion. However we have found that, with rapidly advancing computer power, pre-distortion is currently computationally well-treatable. Further, since in certain cases, computational pre-distortion does not yield overlapping viewing windows for all parts of the hologram (i.e. the fact that a viewing position does not exist where all the hologram image can be observed at once), it is advantageous, for such cases (typically large holograms that are to be viewed with a point source light at close proximity), to use a composite technique comprising both pre-distortion and some deliberate reference beam tracking on recording. Such a composite technique turns out to have significant advantages as it acts to bring the viewing windows of each holopixel into better alignment than when one simply tracks the reference beam such that it matches the replay source.
Pre-Distortion of Data
In the case that we elect to write the hologram with a fixed collimated beam, the image data must be corrected for the disparity between the recording and replay geometries. We shall therefore need to know what information to write on the SIM in terms of the required undistorted data that we want to observe when the hologram is complete. Since current SIM devices are usually fabricated from a fixed and equally spaced grid of pixels, this operation is accomplished by use of the (γ, τ) equations to calculate the required pixel swap transformations.
In the case that an SIM device is used which supports a deformable mesh, then the required distorted image data must be calculated from the given data at viewing using the (ε, σ) equations.
The (ε, σ) Equations
Following our previous analysis equations 204, 206, 207, 208 and 209 may be written as follows:
where we have used R=r, this time and
Clearly the solution of these equations is exactly similar to equations 218 to 223 with y being replaced by a and i being replaced by ε. If (γ, τ) represent a desired SIM image data set at the viewing distance z=hv then the solution of the above equations gives the required distorted dataset (σ,ε) that must be written at the recording distance z=hr.
Recording with a Converging Reference Wave Geometry
In the case that the replay light is too close to the hologram, making it necessary to vary the angle of the recording reference beam from pixel to pixel at production of the hologram, we must re-derive the mathematical model presented above for the case of a general converging reference wave. We will therefore now consider the case of
Equations 179 to 184 need to be supplemented by the following relations:
xh=xr−rrsinθrsinr (232)
yh=yr+rrsinθrcosφr (233)
where
rr2=(xh−xr)2+(yh−yr)2+zr2. (234)
Equations 192 to 194 also need to be supplemented by:
Following 196 to 201 we now define
β=xh−xr (236)
δ=yr−yh (237)
whence equations 232 and 233 yield
sinθrsinφr=−β/rr (238)
sinθrcosφr=−δ/rr (239)
Thus equations 202 and 203 are generalized to
Dividing these two equations we eliminate k1 and k2 as in 204:
Equation 205 is now generalized to:
and 206 to:
Finally with the additional relation
rr2=β2+δ2+hf2 (245)
we can recast our problem in exactly the form of equations 210 to 212.
The Generalized (γ, τ) Equations
As before we may now write
aτ+bγ+gR=0 (246)
where, as before, we have used R=r2 for simplicity. We write the generalized coefficient set as:
These equations may then be solved using the expressions of equations 218 to 223. In
Calculating the Best Choice for the Recording Geometry
In the case of a hologram designed to be lit with a close point source, the above section demonstrated clearly the advantage of using an overcorrected reference geometry on recording. This may be thought of as inducing an effectively benign distortion into our hologram printing system in order to achieve a better viewing situation.
There are various ways of varying the reference beam recording angle in order to achieve better viewing window overlap for a certain replay geometry.
The simplest method is to use an overcorrected reference recording beam of axial angle of incidence equal to that of the intended replay beam but which is characterized by a sink at (xr, yr, zr=hf). Then we calculate hf by preferably minimizing the quantity
Here, Λ (0≦Λ≦Λo)is a coordinate that defines the location of a point on the calculated viewing window perimeter corresponding to a certain holopixel. Its magnitude may either refer to the distance between a chosen reference point, then measuring around the perimeter of the viewing window, to the point in question or it may more preferably refer to the distance around the object window perimeter to the corresponding point (this is consistent as rays transform in a 1 to 1 manor from the object geometry to the image geometry and hence a point on the viewing window perimeter is uniquely associated with a point on the object window perimeter—
The Generalized (ε, σ) Equations
For completeness we note that the generalized (era) equations may be written in canonical form as follows:
where R=r1. The generalized coefficients are calculated exactly as above.
Refinement of the Recording Geometry Choice
We mentioned in the previous section that it is possible to obtain somewhat better viewing window overlap using astigmatic recording beams. In fact we may generalize this to minimizing the function P of equation 254 with respect to individual altitudinal (θr) and azimuthal (φr) reference recording angles corresponding to each holopixel. Equations 238 and 239 may then be used in equations 246 to 248 (or 255 to 257) in order to calculate to required pre-distorted image data.
It should be noted that for the purposes of numerical convergence the following definition of P is, in some cases, preferred:
When the parameter χ=0, this equation is the same as equation 254. When, however, χ=1, however, the minimum of the function P will correspond to the closest situation possible to when all viewing windows are centred on their respective holopixel. By initially choosing χ=1 and then by slowly changing χ towards 0, we can find the minimum of P(χ=0) more easily. In addition, sometimes this minimum does not exist and we may then, using this procedure find a minimum P for a finite and acceptable value of χ.
Various definitions of P may be constructed that are somewhat different to those given above but which serve the same purpose. We do not give further examples here as it will be clear to someone skilled in the art how many types of P function may be formulated (given the known illumination geometry for the hologram) that result in the identification of a specific recording geometry corresponding to optimum hologram viewing characteristics.
Indeed the problem may be cast using various other mathematical formalisms. Each of these formalisms will seek to vary the recording of some or all of the reference angle parameters in order to find a best set of such angles corresponding to the viewing windows of all (or some representative set of) holopixels being generally overlapped better than they otherwise would.
In summary the techniques of Klug et al. search to avoid the necessity of pre-disortion of the image data by exact conjugate matching of recording and replay reference beams. We find that, particularly for the case of large format holograms that are to be replayed with a point source light at close proximity, there is significant motivation for choosing a combination of pre-distortion and reference beam tracking in order to produce better viewing window overlap. This technique may be expected to produce better quality holograms.
In the case of medium format holograms, where the replay light is not so close to the hologram, we find that the mechanically simple solution of a collimated recording beam combined with image pre-distortion provides an optimal solution to the problem.
The Integration of Pixel Swapping Transformations
It has been seen in the first part of this invention how digital image data must be transformed according to special mask transformations in order that this data be in a form in which it can be written to the SLM of the holographic printer. We have also seen how the fundamental distortion of the printer writing objective has to be integrally incorporated in such mask transformations.
In the second part of this invention we have seen how the digital data that we intend to write onto the printer SIM must also be transformed in order to eliminate both the various diffractive distortions inherent to the printer (i.e. recording reference geometry) and by the viewing conditions of the final hologram (i.e. replay reference geometry).
Just as the objective distortion compensation must be integrated into the mask transformation so too must we integrate the diffractive distortion compensation. To see how this works it is best to formally re-discretize the SIM vertical and horizontal co-ordinates in terms of the indices {haeck over (ε)} and {haeck over (τ)} just as we have previously used μ and ν. The reason for this is that we will then retain ν and τ for the description of the projected SIM plane at viewing time.
We now define the tensor {haeck over (ε)}{haeck over (τ)}Wαβ which represents the diffraction compensated data that we intend to write to our SLM for each holopixel (α,β) assuming a paraxial objective. This tensor is related to the paraxial mask tensor S by the following relationship:
{haeck over (ε)}{haeck over (σ)}Wαβ=μνSαβ (259)
Since μ and ν correspond respectively to τ and γ (by a simple linear scaling) we can use equations 218 to 223 in order to write
μ=F({haeck over (σ)},{haeck over (ε)}) (260)
ν=G({haeck over (σ)},{haeck over (ε)}) (261)
where F is the function of equation 218/221 and G is the function of 219/222. Note also that for the generalized case we may use the generalized (γ,τ) equations (this will just change the forms of F and G but otherwise 260 & 261 hold).
Taking the full-parallax max-FOV mask transform of equations 132 to 136 as an example, we may thus write:
Now W must be corrected for objective distortion. If we denote T as the fully corrected mask data then it follows that
whence the integrated transform may be written:
This in then an example of a simple mask transformation that corrects for both diffractive and objective distortion effects. It is clearly a simple matter to apply this logic to any of the previously treated viewing and rendering geometries in order to arrive at other mask transformations incorporating both diffractive disparity and objective distortion effects.
In summary, for large format reflection holograms that are designed to be illuminated by a point source at relatively close proximity the following procedure should be used. A rectangular boundary to the recording SLMI is defined by a series of (ε,σ) coordinates. The function P or equations 254/258 is then minimized, thus defining the preferred recording geometry in terms of the required illumination geometry (i.e. this defines F and G). Then the modified mask transformation (e.g. equations 266d-266h or similar equations describing an integrated mask transformation for finite diffractive disparity and finite objective distortion) is applied to the digital image data in order to calculate the required SLM data. Finally the hologram is written using a recording reference beam tracking that is defined by the calculated optimum recording geometry.
In the case of smaller format holograms where a collimated reference is used on recording, the formalism of transformation 266d-266h is still valid.
The General Case of a Precise Diffractive/Refractive Model
When we study the diffractive and refractive processes in detail that occur at a given holopixel, we see that equations 175 to 178 may generally be replaced by more general expressions of the form
A1(θi, θc, θo, θr, φi, φc, φo, φr, φr, k1, k2)=B1 (267)
A2(θi, θc, θo, θr, φi, φc, φo, φr, φr, k1, k2)=B2 (268)
A3(θi, θc, θo, θr, φi, φc, φo, φr, φr, k1, k2)=B3 (269)
where the functions A1, A2 and A3 are general non-linear functions of the shown variables and the functions B1, B2 and B3 are non-linear functions describing the refractive and emulsion-swelling processes. This leads simply to the functions F and G of equations 260 and 261 being replaced by general non-linear functions. Otherwise the procedure that we have described above remains valid.
We have illustrated the formulation of the generalized mask transformation with the max-FOV case. It should be clear to someone skilled in the art that generally we might optimize this process by variably clipping the final SLM data window such that each and every viewing window corresponding to a given holopixel precisely aligned at replay. These clipping functions are clearly defined after we have minimized the function P.
In addition it should be evident for someone skilled in the art how all the above procedures may be generalized to all the various geometries that have been hereto above discussed.
Discoloration
In the above discussions we have assumed that the swelling parameter α is known and that its effect is compensated for routinely. However, when either swelling is present or the replay geometry of the hologram is not equal to the recording geometry we observe that generally the replay wavelength of a given holopixel is not the same as the recording wavelength.
In the general case of pre-distorted data that we have discussed above, we observe that for each holopixel there is imposed a slightly different discoloration by the applied pre-distortion. This means that the simple application of pre-distortion as described above will lead to a hologram image that appears discolored to different extents depending on holopixel location and on viewing angle.
The resolution to this effect is to calculate the replay wavelength as a function of holopixel location and viewing location and to use this information to modify our colour mixing at each such datum. Generally, for an RGB model, we may thus write
{haeck over (μ)}{haeck over (ν)}Tαβ={haeck over (ε)}{haeck over (σ)}Wαβ=μνSαβ={haeck over (μ)}{haeck over (ν)}UGαβkgIij+{haeck over (μ)}{haeck over (ν)}URαβkgJij+{haeck over (μ)}{haeck over (ν)}UBαβkgKij (270)
where now T is the green mask data, I is the green image data, J is the red image data, K is the blue image data and UG, UR, and UB are respectively functions of the holopixel replay wavelength and the red, green and blue recording wavelengths. Clearly the equations for the blue and red mask data are equivalent.
In some cases the wavelength shift due to the disparity between recording and replay reference beam geometries may become too great for easy correction. In this case the P functions of equations 254 and 258 should be generalized to contain a term proportional to the magnitude of the wavelength shift. Lagrange multipliers may be used to control the importance of these terms within the minimization. Such a constrained optimization will lead to the best window overlap which doesn't unduly discolour the image.
The P functions of equations 254 and 258 may also be modified to contain a term inversely proportional to the final spatial image resolution on the viewing plane. Again Lagrange multipliers may be used to control the importance of this term. Such a constrained optimization would assure the identification of the best window overlap whilst producing a hologram of acceptable peripheral angular resolution.
Number | Date | Country | Kind |
---|---|---|---|
0202139.2 | Jan 2002 | GB | national |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/GB03/00345 | 1/27/2003 | WO | 4/20/2005 |