Synthetic panoramagram

[0002] This application claims benefit of provisional application 60/032,146, filed Dec. 6, 1996.

[0003] The invention is an lenticular stereogram and a method for making the same which provides full panoramagram capability from a stereo pair.

[0004] There is a substantial prior art in the patent and other literature discussing parallax panoramagram or lenticular stereogram technology. It is not our intention to provide a thorough review of the prior art, but rather to provide a background sufficient to orient one skilled in the art enough to fully appreciate the present disclosure.

[0005] Image selection, as Ives has stated, must take place at either the surface of the display or the eyes of the viewer. If the former, then an infinite number of views is required if the image is to be viewed from any location in front of the display. If the latter, then only two views are required. In practice, the interdigitated panoramagram, which is the most practical version of the art under discussion, uses only a handful of images, which may or may not provide acceptable image quality. See L. Lipton, “The Future of Autostereoscopic Electronic Displays”, SPIE Conference Proceedings #1669.

[0006] Initial panoramagram cameras provided optical or mechanical means, with either a moving slit or lenticular array, to create a continuum of views within each image column, but the precursor of the panoramagram, the interdigitated parallax stereogram, used sliced and alternated left and right images of a stereopair. This array may be viewed either through a raster barrier (which resembles a Ronchi grating) or an overlaid lenticular surface. In the case of the lenticular method, each lenticule covers an interdigitated set of images, or one column. For either the barrier or lenticular method, image selection is provided more or less at the plane of the display, and there is no requirement for individual viewing devices or eyewear. The lenticular screen typically has a surface of corduroy-like semi-cylindrical elements or lenticules, and has for the most part supplanted the raster barrier because it has much greater light transmission, and because stereo viewing zones are much larger than unwanted pseudo viewing zones.

[0007] The term “continuum” as used herein is an approximation in fact due to resolution limitations in any imaging medium or lens system, but the term is more descriptive of the methodology of imaging since no after-the-fact combination is required of separately photographed views. The major benefit of this approach, however derived, is that the observer has relatively good freedom of head placement and can see a stereoscopic image over a broad angle of view.

[0008] The interdigitated stereogram, on the other hand, requires exact placement of the eyes in a particular location, and if the head of the observer moves ever so slightly, a pseuodoscopic rather than a stereoscopic image will be seen. The parallax panoramagram is an improvement over the interdigitated parallax stereogram in that it allows for more freedom of view location with respect to the display since quite of bit of head movement is possible. The parallax panoramagram, by incorporating a continuum or multiplicity of views, allows for a wider horizontal angle of view of the display. As previously stated, these cameras are able to produce such a continuum by various means. (The vertical angle of view is not at issue since the lenticular screen is refractive only in the horizontal plane, and head movement up and down does not alter the stereoscopic effect.)

[0009] The panoramagram requires a time exposure in some embodiments. The temporally derived spatial parallax information which is captured during the exposure is transformed into the required spatial parallax information. As long as a time exposure is required, certain applications and the ability to shoot many types of subjects are limited. Moreover, panoramagram cameras are complex and costly instruments. Thus, inventors sought means to simplify the panoramagram process and preserve its best quality, namely that the image is more easily viewed than the more primitive interdigitated stereogram.

[0010] In a clever variation of the two view interdigitated stereogram, inventors created means to interdigitate a multiplicity of views, in some embodiments captured simultaneously with a number of cameras suitably spaced along a horizontal base. Resembling the interdigitated stereogram, the interdigitated panoramagram neatly attempts to combine the best aspects of both techniques. An array of cameras facing the subject is used to simultaneously photograph the subject, and the images are combined optically or with the aid of a computer to produce the necessary columns of information. In some cases, a single camera is preferred for shooting still-life images, and it is slid along a bar for successive exposures.

[0011] It should be noted for clarity that the literature of silver-based photographic panoramagraphy uses the term “interdigitate” to describe the laying down of multiple perspective stripes of images within a column. However, computer graphics people often describe this process as “interleaving,” which is an unfortunate term because it is easily confused with the term “interlacing,” which is widely used to describe electronic displays but with an entirely different meaning.

[0012] The panoramagram and its variants have over the years been used for many applications, such as portraiture, advertising, product packaging and magazine reproduction. The technology has also been used for mass consumer snapshot applications with cameras using only three or four lenses.

[0013] In the past decade or so, products for snapshots have appeared using the interdigitated concept, but using only three or four lenses or views. While better than a interdigitated stereopair stereogram, the result falls far short of a true panoramagram or a decent interdigitated panoramagram with regard to depth effect and viewing angle. Such products have not had a great deal of commercial success in the marketplace.

[0014] In addition to the needs of those requiring still images, viewed either as reflective prints or transparencies, there is also the need for autostereoscopic panoramagram technology as applied to motion picture and electronic imaging displays. The art was applied to movies in the former Soviet Union. Here, a raster barrier was used to view interdigitated stereograms, each view projected with its own lens. After about two decades, the process was abandoned because audiences preferred the polarizing method of image selection since sitting with the head held rigidly was uncomfortable for most people.

[0015] There have been some recent attempts to apply the art to electronic displays. The flat matrix type display panel would, at first, appear to be a good fit because of its ability to provide exact registration of an image element with respect to the lenticular screen optical elements. As it turns out, the commercially available products have been disappointing because of poor image quality and the constraint on viewer head placement, as mentioned above. One way to solve the later problem is to employ head tracking technology to shift columns to maintain a stereo rather than a pseudo effect. However, this technology is costly and can only work for a single observer at a time.

[0016] One of the most important potential applications of the art is for autostereoscopic television, or 3D TV, which may be viewed without eyewear. Laboratory demonstrations have shown that multiple video projectors projecting to the rear of a screen can be viewed from the front through a lenticular array. But even the use of eight or so projectors still results in severe limitations of observer head movement, and obviously requires a large bandwidth if each image requires its own channel.

[0017] One key to making such a television display practical might be to use an approach in which only a stereo pair of images is transmitted, and by some means at the TV set itself, create the additional views required for a true panoramagram display and its attendant benefits. A number of research papers, mostly by workers in Japan and Western Europe, have been presented on this topic. A review of the papers shows that limited success has been achieved, and it is far from certain that a promising approach will be forthcoming because the techniques employed require enormous computational effort. Although a receiver capable of synthesizing a panoramagram out of a stereo pair will require both image storage and powerful computation ability to accomplish on-the-fly synthesis, more progress will have to be shown on the algorithmic level.

[0018] As we have seen, there are significant limitations in the prior art. Progress from the interdigitated stereogram to the panoramagram, and then to the interdigitated panoramagram, is significant. However, production of still images with a multiple camera/lens array is not practical for many applications such as consumer snapshot applications or motion picture cinematography. In addition, there are many workers using field sequential stereo displays on computer workstations (for example, for scientific visualization) and viewing the display with shuttering eyewear, and these workers have a need for hard copy.

[0019] There is no question that a continuous approach, as illustrated in terms of individual columnar structure, is far superior to any other method in terms of ease of viewing and depth effect. Thus, it would thus be highly desirable to produce a full panoramagram effect from a stereopair, since the aforementioned applications would become practical. There would be no need for a multiple lens/camera array and the ensuing complications for still photography. It is furthermore clear that the usual panoramagram is not a practical solution for motion pictures, which would preferably be carried out with at most two cameras for the production of a stereo pair. Moreover, projection would then be most conveniently carried out with but a single projector.

[0020] The present invention includes a panoramagram and a method for making the same from a pair of planostereoscopic source images. The pair of source images are, for example, a left image and a right image having different perspective views which are spaced apart in a horizontal plane. The left image may be considered the starting image and the right image the ending image, or vice versa. Control points are defined at corresponding locations in the source images, and each control point includes position and color information. A plurality of intermediate images are created by “morphing” from the starting image to the ending image using position and color information from each of the corresponding locations. Preferably, the morphing process involves creating transformations of the source images based on the physical proximity from each control point; in other words, the transformation is affected strongly by control points which are nearby and is less affected by control points which are far away. The intermediate images and the source images are then interdigitated to create a single output image with a continuum of views ranging from the left image to the right image.

[0021] A better understanding of the features and advantages of the present invention will be obtained by reference to the following detailed description of the invention and accompanying drawings which set forth an illustrative embodiment in which the principles of the invention are utilized.

[0022] FIG. 1A is a perspective view showing a lenticular screen and a display surface in intimate juxtaposition according to the prior art.

[0023] FIG. 1B is a perspective view of an individual lenticule and its associated image column according to the prior art.

[0024] FIGS. 2A, 2B and 2C are plan views of a classical stereo continuum panoramagram, an interdigitated stereopair panoramagram, and a multiple view interdigitated panoramagram, respectively, all according to the prior art.

[0025] FIG. 3 illustrates the process flow for creating a synthetic panoramagram in accordance with the present invention.

[0026] The object of this invention to create a continuous panoramagram synthetically from just two views, namely a left view and a right view. By beginning with an ordinary stereopair, it is our goal to produce the same excellent depth effect that has been available, until now, only by means of the classical panoramagram. The taking of a panoramagram image is difficult and has heretofore been limited in its application to static images.

[0027] In addition, we have achieved the ability to create not simply an interleave or interlaced effect, but rather the continuous panoramagram effect, which has vastly superior image quality.

[0028] There are two independent but related parts of the disclosure. One relates to the ability to begin with a stereo pair and to produce from that pair of images the interdigitated panoramagram. The second innovation relates to software method for taking these individual interdigitated elements and turning them into the continuum needed to produce an intrinsic panoramagram.

[0029] FIG. 1A shows a typical lenticular sheet 103, which is made up of individual lenticules 102 arranged in intimate juxtaposition with a display surface 101. The display is arranged in columns, as shown in FIG. 1B, wherein column 104 is viewed through lenticule 105. Each column has a recording of perspective information with the left image information on the right side of the column and vice versa, as illustrated in FIGS. 2A through 2C. This inversion of left and right results from the optics of the lenticular screen.

[0030] With reference to FIG. 2A, column 201 illustrates the intrinsic panoramagram which uses a recording of a continuum of views, as represented by a horizontal line of pixels 202. The resolution of pixel density along this horizontal line determines the number of individual views which may be recorded in a column and thus viewed through a lenticule.

[0031] With reference to FIG. 2B, the column 203 is divided into two image stripes, namely the right perspective view 204 and the left perspective view 205. This is the structure of an interdigitated parallax stereogram. It has the limitation described in the Background section above and requires exact placement of the observer with respect to the display surface.

[0032] FIG. 2C is an interdigitated panoramagram constructed out of many views, with column 206 including right extreme view 207, middle view 208 and left extreme view 209. Although superior to the interdigitated stereogram shown in FIG. 2B in terms of head movement, the observer will still be aware of discrete “flips” or discontinuities because of the discrete nature of the image structure.

[0033] FIG. 3 shows the steps used to create a synthetic panoramagram in accord with the present invention by using any stereo pair as the source. Processing is presented into four basic stages, namely (1) image preparation, (2) pattern recognition, (3) morphing and (4) interdigitating.

[0034] A brief overview of the inventive technique follows. Stereo image pair preparation (1) is the standard means for creating plano-stereo images (two view images which are synthesized by the mind into an image with binocular stereopsis) for any medium, such as silver-based photography, digital photography, or computer generated images. For example, one can start with digitized left and right stereo files 10L and 10R in a Windows Bitmap (BMP) 24 bit RGB format. If files are in any other format, convert as outlined using any off-the-shelf image program, such as Photoshop, Photosyler, or Picture Publisher. Make alignment adjustments (rotation and vertical) as you would for viewing stereo images with any medium. Set the window following the basic rules as outlined in the StereoGraphics Developers Handbook, Chapter 3: Composition, available on the internet at www.stereographics.com.

[0035] Pattern recognition (2) can be accomplished in two ways, either by means of a pattern recognition algorithm as illustrated in block 20, or by manually selecting similar location points in the digitized stereo pair using an off-the-shelf morphing program. Morphing, or “transition morphing” as it is called in .this case, is the transformation of an image to form a new image using position and color information from both source images. One defines control points in each image such as start point 20aand end point 20b. These two control points specify an exact amount of movement for the point of the image which is exactly under that particular control point. Areas of the image which are not directly underneath a control point move in such a way as to agree with all the control points nearby; the closer a control point is to a portion of the image, the more the image will move in a manner similar to the motion of that point. Control points which are very far away from a location on an image will have little or no effect on that location.

[0036] The term color information as used herein is fully intended to encompass monochrome data, such as image density information.

[0037] Morphing (3) is the generation of a sequence of images to provide a continuous change from one image to the other, such as illustrated in block 30. Determining the proper number of views for a particular lenticular screen usually is dictated by the physical width of a lenticule (1/pitch), that is, how many printed pixels will actually fit under a lenticule. It is possible to improve the quality of depth and field of view by increasing the number of views, applying an algorithm that averages the x number of views, and resize the x number of views to place under the lenticule.

[0038] Next, the step-by-step method we have used is given in some detail explaining the exact combination of techniques used to make a print to be viewed with a lenticular screen using any stereo pair as the source.

[0039] Stage 1—Image Preparation

[0040] The first four steps prepare images for the morphing process and are conventional for preparing stereo images for any medium.

[0041] Step 1. Start with digitized left and right stereo files. A TIFF (Tagged Information File Format) format is preferably be used, and converted to RGB. If files are in any other format, they may be converted using any suitable image program, such as Photoshop, Photosyler, or Picture Publisher. One must be sure to use the same image program to convert both the left and right image files because different image programs can save files with different header information which will corrupt the interdigitation process.

[0042] Step 2. Make alignment adjustments (rotation and vertical) as one would for viewing stereo images with any medium, for example, using the Photoshop program. Picking similar X,Y points on each of the left and right image files will provide a good reference for calculating the amount of correction required for both vertical and rotational adjustment. It is best to adjust rotational error first, then vertical alignment second.

[0043] Step 3. Set the window or zero parallax point by horizontally shifting the images with respect to each other following the basic rules as outlined in the StereoGraphics Developers Handbook, Chapter 3: Composition. The Photoshop program may be used to get an X,Y point on each file of the object that will be placed “at the window,” then crop each file to match the selected X,Y points.

[0044] Step 4. Image resizing is predetermined by the number of views to be made. For example, if 20 views are used for a lenticular presentation, then the width of each image file must be divisible by 20. A good working size is between 1000 and 2000 pixels wide maintaining the proper aspect ratio for the height. As an example, suppose the image files are 3023 pixels wide by 2267 pixels high after making alignment adjustments and setting the window, and one wants 20 views for a lenticular print. Using the Photoshop program, load the left image file, select Image, Image Size dialog, type in 2000 pixels (divisible by 20) in the width box and press OK. Photoshop will maintain the proper aspect ratio for the height (1500 pixels) if the proportions box is checked. Then do the same for the right image file. Save the files. The left and right image files are now ready for the morphing process.

[0045] Step 5 (optional). Using Photoshop, duplicate left/right images and apply an edge/invert filter to each image to define distinguishing landmarks and edges for a more accurate point selection in the morphing process. This process only works on rather high contrast images and proves less effective with low contrast or monotone images.

[0046] Stage 2—Morphing

[0047] As previously mentioned, morphing is the transformation of one or both images to form a new image, using position and color information from both source images. Positional control may be determined by manual placement. Morphing is the generation of a sequence of images to provide a continuous change from one image to the other. One places control points, which actually consist of two elements, namely start and end points. These specify an exact amount of movement for the point of the image which is exactly under that particular control point. Areas of the image which are not directly underneath a control point move in such a way as to agree with all the control points nearby; the closer a control point is to a portion of the image, the more the image will move in a manner similar to the motion of that point. Control points which are very far away from a location on an image will have little or no effect on that location.

[0048] There are several off-the-shelf morphing programs available which can adequately perform this task, for example, Elastic Reality and WinImages:Morph. The latter program has the distinct advantage since it builds a separate ASCII file with the start/end control points as X,Y position relative to a master grid. This file could be built by other programs, such as a pattern recognition program (PRP) to automate the control point selection. It is our expectation that this process can be automated by use of such PRP techniques, to be applied to our process at some future time. We describe the morphing process using WinImages:Morph for the conversion of stereo image pairs to lenticular output.

[0049] Step 6. In order to create a transition morph, the left image is loaded as the start frame and the right image is loaded as the end frame. For example, if the stereo image pair is a portrait scene, there will be two color image views, one of the start image and one of the end image. These two image views act like any other window in that they can be re-sized and re-positioned to best suit specific needs. These images are used as a guide in placing the control points.

[0050] Step 7. Re-position the start frame to the right on the screen and the end frame to the left on the screen. This allows cross-viewing of the stereo images when selecting control points. It is not necessary for selecting all control points, but some points are difficult to place without the cross-viewing technique.

[0051] Step 8. Select the Sequence Controls option in the Generate menu. When the dialog opens, check the text entry field marked Total Frames. If it does not read the value 20, then click in this field with the left mouse button and replace whatever was there with 20. (For this example we are using 20 to be consistent with our file width of 2000 pixels.)

[0052] Determining the proper number of views for a particular lenticular screen is usually dictated by the physical width of a lenticule, that is, how many printed pixels will actually fit under a lenticular lens. Experience has shown it is possible to improve the quality of depth and field of view by increasing the number of views, applying a algorithm that averages the x number of views, and resize the x number of views to place under the lenticule. More on this in Stage 3.

[0053] Step 9. Zoom in on the working area for the left eye. To zoom, first click on the zoom button located in the top right-hand corner of the morphing panel. Notice that the normal WinImages pointer will change into the shape of a magnifying glass. Now point the zoom pointer about ½ inch above and to the left of the left eye (start image), click and hold the left mouse button and drag the mouse down and to the right. A box will appear; keep pulling until the box entirely encloses the eye (about ½ inch below and right) and then let go of the mouse button. The box can be repositioned by pressing the right mouse button in addition to the left. The image will now be zoomed in.

[0054] If the result is not pleasing, click on the dezoom button and the zoom will be turned off. Repeat these steps, beginning with re-selecting the zoom button, to get a better zoom around the eye. When the eye is nicely centered in a zoomed image, continue. The image in the right (end) view will need to be zoomed also. Follow the steps above to zoom around the left eye in the end image view.

[0055] Step 10. Select the point tool from the tool box, then point at the eye of the start image and drop points all around the edge of the eye (about 8 points should be just fine). Preferably, the points are placed as follows; one at each corner of the eye; one directly above, and one directly below the pupil; and the final four, one at a time between the corners of the eye and the points by the pupil.

[0056] Step 11. Looking at the end image, the same points will be visible, but they will not be correctly arranged around the end image's eye because they are in different X positions due to placement of the window and the fact they are different views of the same object. In the manual process, those points must be arranged in the same manner around the other eye. Obviously, the points should be in the same relative positions as they are in the left frame. Point and move the points in the right frame by right clicking on a point and then moving it into position. If one has the ability to cross-view the images, one can place a control point correctly by using depth cues. If the need exists to adjust any of the points in the start image view, right click on the desired point and move it to its new position. Adding more points will dramatically improve the details in the morphing process.

[0057] Step 12. Continue as outlined in steps 9 through 11 on other regions of the image until all portions of the image are covered with control points.

[0058] Step 13. Save the project. This step will generate two files, a project file with image names and directory locations, and a control point ASCII database listing all the X,Y coordinates of the start frame and end frame.

[0059] Step 14. Select the Sequence Controls option in the Generate menu. This is where the user sets the working parameters for the morph. These include resolution, output format, and number of frames. The descriptions below detail all of these dialog controls.

[0060] Total frames: The total number of frames in the morph, set to 20 in this example.

[0061] Begin Frame #: The frame number where morphing will begin for a sequence generation. Set to 1.

[0062] Finish Frame #: The frame number where morphing will finish for a sequence generation. Set to 20.

[0063] Current Frame #: The frame that will be created when “Do one Frame” is selected. For example, if Total Frames is set to 20, a Current Frame of 10 will generate the middle frame, half way between the Begin and Finish frames. If you then set Begin Frame to 1 and Finish Frame to 20, all the frames of the morph animation will be generated.

[0064] Anti-Alias: Anti-aliasing affects the spatial (or positional) accuracy of the morph result in distances less than one pixel width. This should be switched on for the best quality result.

[0065] Save Result: When a morph frame is generated, this switch tells the program to automatically save the resulting image. The files will be saved in the format specified in the Save File as Type option found in this dialog. Set to TIF format.

[0066] Output file: The file name that will be used when the morph result is automatically saved.

[0067] Specify: This opens a file requester to locate a valid path and file name for the output file. The last four characters of the file name (before the extension) will be replaced with the frame number (0001 to 9999). For example, a file having the name C:\IMAGES\MYIMAGE.TIF will be changed to C:\IMAGES\MYIM0007.TIF for frame seven.

[0068] This requester also contains all of the output file information, including format, Color Depth, Compression, and output path and file name. Morph will automatically alter the output file format if you change the extension of the image name.

[0069] Save File as Type: This drop down box allows you to select the output file format for a sequence of morph frames. Changing the save type will also change the save options (Color Depth, Compression, and JPEG Quality) that are available. For example, if you select BMP as the output format, you would be able to select a 24 bit, 8 bit, 4 bit, or 1 bit render Color Depth for that format.

[0070] Output size: This is the resolution of the results of morph creation. It is the main determinate of the output quality. This should be set to match the original size of the left/right image.

[0071] Match Start Aspect Ratio: Matches the output size to the ratio of dimensions of the Start Frame.

[0072] Match End Aspect Ratio: Matches the output size to the ratio of dimensions of the End Frame.

[0073] Set to Start Frame size: Matches the output size to the exact dimensions of the Start Frame.

[0074] Set to End Frame size: Matches the output size to the exact dimensions of the End Frame.

[0075] Step 15. Select the Sequence Generate option in the Generate menu. This generates an entire sequence of morph frames. The extent of the sequence was specified in the Sequence Controls and usually goes from Frame #1 to the last frame (which equals the Total Frames value set in the Sequence Controls). If the Save Result option was set in the Sequence Controls, then each frame will be saved as it is generated.

[0076] Stage 3—Interdigitating Process

[0077] We have developed a program which vastly speeds up the task of interdigitating the multiple views. In our preferred embodiment, the program is written in C language and executed on a Pentium-based personal computer with Windows NT operating system. A copy of the program is set forth in Exhibit A. The program is so much faster than prior art methods that one may use a computer of ordinary capability rather than one of extraordinary performance. We have found the program given here produces lenticular stereograms free from disturbing artifacts which detract from the three-dimensional effect. In addition, it is desirable to use the program to compute more views than calculation predicts can be resolved behind each lenticule. By this means we have achieved the continuous panoramagram effect, as opposed to the interleaved effect, which has noticeable visual jumps as one moves his or her head within the acceptable viewing zone. Moreover, the stereoscopic effect of our new technique is much deeper, and there is much less blurring of image points away from the zero parallax point, a common complaint reported in the literature.

[0078] For example, if calculation predicts that the pixel limit along a horizontal line within a column is 20, we find that by providing at least 30 views but not more than 80, we have achieved the continuous panoramagram effect with its attendant benefits. Experience has shown that between 1.5 and 4 times the number of pixels predicted as the resolution limit will have the beneficial effect sought.

[0079] In addition, the method described here has the benefit of providing a final image output to any print size with any number of views of almost any size, thus providing a wider range of print size choice. Additionally, a special border is calculated and added to the final output for the purpose of precision alignment under a lenticule.

[0080] The program reads any number of bitmap views, vertical line by vertical line, adds additional border lines, and places the extracted data homogeneously into an anamorphic sized output file.

[0081] If the calculations require the anamorphic file to be larger than the total sum of the vertical lines extracted from the total views, a process of stretching is applied. Stretching is the process of adding “holes” evenly across the anamorphic file, then filling the holes with average interpolated pixel data.

[0082] If the calculations require the anamorphic file to be smaller than the total sum of the vertical lines extracted, then vertical lines are “dropped” homogeneously across the anamorphic file.

[0083] Advantageously, the image data can be handled one scan line at a time. Vertical lines from each of the views can be extracted in two methods. Under the first method, all vertical lines are extracted from the same pixel position of every view. The second method extracts vertical lines from every view sequentially within a given “block.” A block is calculated as (number of views*lens pitch*print inches)/(number of views). Either method will give the desired result once placed under the lens, but the raw data will look quite different, especially that image information in the area where the stereoscopic information is “at the stereo window,” or at the “zero parallax point.” The anamorphic file is then resized to the proper final size using an average interpolation algorithm.

[0084] Step 16. Add special border to views for alignment purposes. We add a black border to the first two images and a white border to the rest.

[0085] Step 17. Interdigitate the views and make one file for output to printer. Interdigitating may be accomplished using conventional programming methods. We have developed a suitable interdigitating program using a Visual Basic program, a C++ program, and the Photoshop program, as shown in Exhibit A. The basic process involves resizing each view based on number of views, grabbing a vertical line from each view and placing it in the final output file, and resizing the output file for proper output.

[0086] We have described in this disclosure means for producing lenticular stereograms of the panoramagram type from a traditional stereo pair of images. The means we have are given with exacting specificity and detail, and because of this we caution that others may find related means to achieve the same effect, and we remind the reader that the fundamental principals are of importance in disclosing our invention. Others skilled in the art may uncover useful variations, but only after having been apprised of the basic technique.

[0087] The invention may be applied to so-called hard copy of the reflective or transmissively illuminated types, and it may be applied to slide and motion picture projection. In addition, it may be used as a bandwidth conserving technique when applied to electronic imaging, such as television and computer graphics, and there may be display applications we have not imagined.

[0088] Although the technique has been described in terms of display with a lenticular screen, only a portion of the techniques given herein relate specifically to that selection medium. Autostereoscopic displays in general depend on having multiple images viewed in conjunction with some type of selection device, such as a lenticular screen. The basic morphing technique for synthesizing multiple views may be applied to other selection techniques using various raster and lens arrangements, for hardcopy, projection, computer graphics display, or video. Indeed, the technique has important applications in holography since at last, a stereo pair produced with a conventional two lens stereoscopic camera may be used to produce the information required in a holographic display. Further, the technique for handling image data one scan line at a time provides a significant advantage in speed of processing the data, and therefore has obvious applicability to real-time data processing. For example, the pixel transfer process will generally be done in an ascending order which can be readily corresponded using computer techniques to real time processes.

[0089] It should be understood that the invention is not intended to be limited by the specifics of the above-described embodiment, but rather defined by the accompanying claims.