Stereo image generating device, stereo image generating method, and program

Abstract

A stereo image generating device according to the disclosed technology generates a stereo image from image information, a parallax map of the image, and viewpoint information of an observer, the image generating device including a parallax remapping function generating unit, a multiband parallax map generating unit, a parallax induction pattern generating unit, and an image pair generating unit. The parallax remapping function generating unit generates a remapping function for converting a parallax amount to a parallax amount within a predetermined range. The multiband parallax map generating unit generates a parallax map in which a multiband parallax map has been corrected with the parallax remapping function. The parallax induction pattern generating unit generates an image in which a bandpass-decomposed image is π / 2-shifted with a spatial phase, and generates a parallax induction pattern by weighting the multiband parallax map with a value and adding the weighted multiband parallax map. The image pair generating unit generates a stereo image pair by adding and subtracting the image information and the parallax induction pattern.

Description

[Technical Field]

[0001] The disclosed technology relates to a technology for generating stereoscopic images to be presented on a stereoscopic television (3DTV) or the like in accordance with the viewpoint position of a viewer. [Background technology]

[0002] When observing the outside world in a natural environment, the human eye performs convergence eye movement and lens accommodation in tandem. Specifically, when gazing at an object in a 3D environment, the eyeballs rotate to minimize the parallax of that object on the retina in the central visual field, and at the same time, the lens adjusts its focus so that the object is in focus. This ensures that the focal length and the point where the lines of sight of the left and right eyes intersect (convergence distance) are consistent, regardless of the depth position of the object being gazed at. Figure 1, "Natural Observation 101," shows what happens when the focal length and convergence distance are consistent.

[0003] However, when viewing 3D images displayed on a 3D TV, for example, the relationship between this convergence eye movement and focus accommodation is disrupted, and while the convergence distance changes according to the parallax of the displayed image, the focal length is always fixed to the display screen, which is thought to cause fatigue and discomfort when viewing 3D images. Figure 1, "Stereo 3D Viewing 102," shows what happens when the convergence distance is shorter than the focal length.

[0004] To solve this problem, a method has been devised in the past to measure the viewpoint position of the 3D observer and remap the parallax of the presented image based on the measured viewpoint position (Non-Patent Document 1). This method dynamically manipulates the parallax so that the parallax near the viewpoint position of the 3D observer coincides with the screen surface, thereby creating a state in which both the convergence distance and focal length are maintained at the distance to the screen. Figure 2 shows a schematic diagram of this process. The observer's viewpoint is measured, and if the viewpoint is on a rectangular parallelepiped, the parallax related to the rectangular parallelepiped is adjusted to near zero, and the parallax of figures other than the rectangular parallelepiped is compressed. If the viewpoint moves to a cylinder, the parallax related to the cylinder is adjusted to near zero, and the parallax of figures other than the cylinder is compressed.

[0005] Another problem with 3DTV is that when an observer without 3D glasses views a 3D image, the left and right stereo images appear to overlap, causing blurring or double images in the image. To solve this problem, Non-Patent Document 2 devised a stereo image generation technology (hereinafter referred to as Hidden Stereo) that does not produce double images when the left and right images are added together. Figure 3 shows a schematic of Hidden Stereo. If the left image (L = I + D) and the right image (R = ID) can be separated using a 3D negative, a 3D image can be observed. If glasses are not required and L and R cannot be separated, a 2D image (L + R = 2I) with the parallax information canceled can be observed. [Prior art documents] [Non-patent literature]

[0006] [Non-Patent Document 1] P. Kellnhofer et al., “GazeStereo3D: seamless disparity manipulations,” ACM Trans. Graph., vol. 35, No. 4, Article 68, pp.1-13, Jul. 2016. [Non-patent document 2] T. Fukiage et al., “Hiding of Phase-Based Stereo Disparity for Ghost-Free Viewing Without glasses,” ACM Trans. Graph., vol. 36, no. 4, pp. 1-17, Jul. 2017. Summary of the Invention [Problem to be solved by the invention]

[0007] Viewpoint-based parallax remapping technology is limited to a single 3D observer because it requires measuring the viewpoint of the 3D observer. For a 2D observer, the image quality deteriorates further as the 3D observer moves their viewpoint, resulting in blurring and double images. Therefore, applying viewpoint-based parallax remapping technology to 3DTV limits the number of observers to one, completely eliminating the advantage over goggle-type displays of being able to be enjoyed by multiple people in the same place. Additionally, Hidden Stereo has a limit to the amount of parallax it can provide, at roughly 6-8 minutes of visual angle. Normal stereo images contain about 1-2 degrees of parallax (1 degree = 60 minutes), so to fully reproduce this parallax in Hidden Stereo, the parallax must be compressed. However, simply compressing the parallax makes it impossible to express subtle differences in depth, resulting in a flat image. [Means for solving the problem]

[0008] In order to solve the above problem, the stereo image generating device of the disclosed technology is a device that generates a stereo image from image information, a disparity map of the image, and observer viewpoint information, and includes a disparity remapping function generating unit, a multiband disparity map generating unit, a disparity inducing pattern generating unit, and an image pair generating unit. The parallax remapping function generator generates a remapping function that converts the amount of parallax of pixels within a certain range from the viewpoint into an amount of parallax within a predetermined range. The multiband disparity map generator generates a first multiband disparity map by low-pass filtering the disparity map, and generates a second multiband disparity map by correcting the first multiband disparity map with a disparity remapping function. The parallax inducing pattern generating unit generates a first band-pass image by band-pass decomposing the image information, generates a second band-pass image by shifting the first band-pass image by π / 2 in spatial phase, and generates a parallax inducing pattern by weighting and adding the second band-pass image with the value of the second multi-band parallax map. The image pair generator generates a stereo image pair by adding and subtracting the image information and the parallax inducing pattern. [Effects of the Invention]

[0009] The disclosed technology combines viewpoint-based parallax remapping with Hidden Stereo to create a system in which the two technologies complement each other's weaknesses. Specifically, Hidden Stereo allows multiple 2D viewers to enjoy the video simultaneously, thereby compensating for the weakness of viewpoint-based parallax remapping technology, which limits the number of viewers to one. Meanwhile, viewpoint-based parallax remapping technology improves the sense of depth provided by Hidden Stereo by remapping the amount of parallax near the viewpoint to within 8 minutes, which is the limit of Hidden Stereo. Fusion is the process by which humans fuse retinal images between the two eyes in the brain without blurring. It is known that the maximum amount of parallax that can be fused is approximately 10 arcminutes (Panum's fusion area). Under natural viewing conditions, the amount of parallax near the point of gaze is concentrated at around 0 due to convergence eye movement, so this fusion limit is often not a problem. The results of the parallax remapping provided by this invention do not deviate significantly from the parallax conditions found in such natural environments, providing a relatively natural viewing experience. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 2 is a diagram illustrating focal length and convergence distance. [Figure 2] FIG. 1 is a diagram illustrating the conventional technology "remapping." [Figure 3] FIG. 1 is a diagram illustrating the conventional technology "Hidden Stereo." [Figure 4] FIG. 1 is a diagram illustrating Equation 1. [Figure 5] 10A and 10B are diagrams illustrating a positive parallax amount and a negative parallax amount. [Figure 6] FIG. 1 is a functional block diagram of a stereoscopic video generating device according to a first embodiment. [Figure 7]FIG. 10 is a flowchart of the operation of a disparity remapping function generation unit. [Figure 8] FIG. 10 is a diagram illustrating a disparity remapping function. [Figure 9] FIG. 10 is a flowchart of the operation of the multiband disparity map generator. [Figure 10] FIG. 10 is a flowchart of the operation of the parallax inducing pattern generating unit. [Figure 11] FIG. 10 is a flowchart of the operation of the image pair generation unit. [Figure 12] FIG. 10 is a functional block diagram of a stereoscopic image generating device according to second and third embodiments. [Figure 13] FIG. 10 is a flowchart of the operation of a disparity map generation unit. [Figure 14] FIG. 11 is a flowchart illustrating the operation of a multiband disparity map generation unit according to the third embodiment. [Figure 15] FIG. 2 is a diagram showing an example of the functional configuration of a computer. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, embodiments of the disclosed technology will be described in detail. Note that components having the same functions are assigned the same numbers, and duplicated descriptions will be omitted.

[0012] In what follows, we use the visual angle (minutes) as a unit of measurement to specify parallax and the range in an image, but this can be converted to pixel units using the viewing distance O (cm), screen size S (cm), and the number of pixels on the screen W. For example, converting visual angle M (minutes) into a pixel unit value P is as shown in Equation 1 in Figure 4.

[0013] Furthermore, "parallax" is given by the distance between the intersection of the left eye's line of sight and the screen and the intersection of the right eye's line of sight and the screen when an object displayed on a screen is viewed at the center of both eyes. This is shown in Figure 5. Parallax can take on positive and negative values. A positive parallax value represents a depth in front of the screen surface, while a negative value represents a depth behind the screen surface. A parallax of 0 represents a depth equal to the screen surface. The parallax values ​​for each pixel arranged according to the image is called a parallax map.

[0014] Hereinafter, a first embodiment will be described in which one 2D image and a corresponding disparity map are input, and a second and third embodiments will be described in which one stereo image pair is input.

[0015] [First embodiment] FIG. 6 is a functional block diagram of a stereo image generating device 601 according to the first embodiment. The disparity remapping function generator 602 generates a disparity remapping function g from the disparity map D(x, y) and the viewpoint position (x, y). The multiband disparity map generator 603 generates a multiband disparity map G by remapping the disparity map decomposed for each spatial frequency band using a disparity remapping function g. D f Generates (x, y). The parallax-inducing pattern generator 604 generates a 2D image I(x, y) and a multi-band parallax map G D F From (x, y), the induced disparity pattern I D Generates (x, y). The image pair generator 605 generates a 2D image I(x, y) and a parallax-inducing pattern I D From (x, y), a hidden stereo image pair I L (x,y) and I R Generate (x,y).

[0016] <Generating a parallax remapping function> This is applied to the disparity map to shift the disparity near the viewpoint to zero, and generates a "disparity remapping function" to compress the amount of disparity to a range that can be reproduced in Hidden Stereo. FIG. 7 is a flowchart illustrating the operation of the disparity remapping function generation unit. The disparity remapping function generator first obtains the disparity map D(x, y) and the current viewpoint position (x, y) in the image (step S701). The disparity amount of pixels within a visual range of 2.5 degrees from the viewpoint position is extracted from the disparity map to construct a histogram (step S702). The minimum value min(D) and the 5th percentile p 05 , 95th percentile p 95 , the maximum value max(D) is calculated (step S703). The minimum output value d of the disparity remapping function min and maximum output d max Using the above, the control points P1, P2, P3, and P4 are determined as follows (step S704 and FIG. 8). P1=(min(D), dmin) P2=(p05, dmin) P3=(p95, dmax) P4=(max(D), dmax) where d min and d max is the disparity range that can be reproduced in Hidden Stereo, for example, d min =-8 minutes, d max =+8 minutes and give it as a constant.

[0017] The most frequent value of the extracted disparity amount p 50 In order to make the output disparity 0 at (50th percentile), the line connecting P2 and P3 is at point (p 50 , 0) (step S705). At this time, the output value of the control point is d min and d max The part beyond the range of d min and d max may be clipped so as not to exceed The control points are smoothly interpolated using the Piecewise Cubic Hermite Interpolating Polynomial (PCHIP) method to obtain a disparity remapping function g (step S706). The disparity remapping function generated by the above procedure is shown in Figure 8.

[0018] <Generating a multi-band disparity map> To generate Hidden Stereo images according to the viewpoint position, it is necessary to speed up the Hidden Stereo image generation process. For this reason, the "multi-band parallax map generation unit" and "parallax induction pattern generation unit" of the disclosed technology perform conversion based on one-dimensional image processing that takes into account only horizontal spatial frequency information, rather than conversion based on two-dimensional image processing that takes into account spatial frequency and orientation as proposed in References 1 and 2 below. [Reference 1] T. Fukiage et al., "Hiding of phase-based stereo disparity for ghost-free viewing without glasses," ACM Trans. Graph., vol. 36, No. 4, Article 147, pp.1-17, July 2017. [Reference 2] JP 2018-56983

[0019] Now, we generate a multi-band disparity map to be used in the disparity induction pattern generation section. f For each spatial frequency band, the peak spatial frequency ω f Parallax map G with the following roughness D f (x, y), where f represents the index of each spatial frequency band. f can be determined according to the number of pixels W in the horizontal direction of the input image. For example, N f =ceiling(log2W-3). ceiling(x) is the ceiling function, which gives the smallest integer greater than or equal to the variable x.

[0020] FIG. 9 is a flowchart showing the operation of the multiband disparity map generator. First, the disparity map D(x, y) and the disparity remapping function g are obtained (step S901). Next, for the disparity map D(x, y), a horizontal moving average is calculated for each spatial frequency band f within a neighborhood window range with the number of pixels corresponding to that wavelength (band-wise low-pass filtering). To reduce the amount of calculation, the moving average processing is performed independently for each horizontal scanning line, and the results are combined to obtain G' D f (x, y) is obtained (step S902). Then G' D f The disparity of each pixel of (x, y) is converted through the disparity remapping function g, and G D f (x, y) is obtained (step S903).

[0021] <Generation of parallax-induced patterns> 3DTV responses are typically not linear, and to offset this, input images are often encoded in formats such as sRGB. In this case, the image is converted from sRGB to a linear RGB space as a preprocessing step. In the following process, the same processing is performed independently for each RGB channel. However, to reduce the amount of calculation, a disparity induction pattern can be generated based only on the luminance (Y) channel after converting from RGB to a color space such as YUV.

[0022] FIG. 10 is a flowchart showing the operation of the parallax inducing pattern generating unit. First, we take a 2D image I(x, y) and a multi-band disparity map G D f (x, y) is acquired (step S1001). The subsequent processing is performed independently for each horizontal scanning line of row y in the image, so I, G D f , A f The coordinate y is omitted. I(x) is subjected to a one-dimensional discrete Fourier transform, and a bandpass filter Ψ corresponding to each spatial frequency band f is generated. fThis bandpass filter is a one-dimensional version of the Complex Steerable Pyramid. However, since the spatial frequency bands corresponding to the high-pass residual component and the low-pass residual component are not used, the corresponding filters are not used. After that, a discrete inverse Fourier transform is performed to obtain a one-dimensional bandpass image B for each spatial frequency band f. f (x) is obtained (step S1002, bandpass decomposition).

[0023] B f By extracting the imaginary component of (x), the spatial phase shifted component ~B by 90 degrees (π / 2) is obtained. f (x) is obtained (step S1003). ~B f Weight A to be applied to (x) f (x) is calculated using the following formula 2 (step S1004). A f (x)=tan(ω f G D f (x) / 2) (Formula 2) To prevent the weight from becoming too large when the disparity is large, A f (x) is clipped according to the following formula 3, and A' f (x) is obtained (step S1005). A' f (x)=min(max(A f , -1), 1) (Equation 3) Note that max(a, b) is a function that returns the larger value of a and b, and min(a, b) is a function that returns the smaller value of a and b. Finally, A' f The 90 degree (π / 2) phase shift component (A' f (x)~B f (x)) is subjected to a discrete Fourier transform and a bandpass filter Ψ f Applying the discrete inverse Fourier transform, the disparity induction pattern I for each scan line is obtained. D (x) is obtained (step S1006). I D (x) is connected in the y direction to form the parallax induction pattern I. D(x, y) is obtained (step S1007).

[0024] <Generation of Hidden Stereo Image Pair> FIG. 11 is a flowchart showing the operation of the Hidden Stereo image pair generation unit. First, the 2D image I(x, y) and the disparity induction pattern I D (x, y) is acquired (step S1101). Next, the left image I L (x, y) = I(x, y) + I D (x, y), and the right image I R (x, y) = I(x, y) - I D (x, y), a Hidden Stereo image pair is generated by pixel-by-pixel addition and subtraction processing (step S1102).

[0025] When generating the disparity induction pattern only for the luminance channel, after performing the above processing on the Y channel, it is converted to the RGB color space. After that, if necessary, the image is converted from the linear RGB space to the sRGB space and then output to the 3DTV. When the range of displayable pixel values is exceeded by the addition and subtraction processing, the components exceeding this range can be calculated and removed from I D (x, y) in advance by clipping processing.

[0026] [Second Embodiment] The case where a stereo image pair is given as the input image will be described. FIG. 12 is a functional block diagram of the stereo image generation apparatus 1201 according to the second embodiment. The difference from the first embodiment is that the input is a stereo image pair and it includes a disparity map generation unit 1206 that generates a disparity map from the stereo image pair.

[0027] <Generation of Disparity Map> FIG. 13 is a flowchart for explaining the operation of the disparity map generation unit. The disparity map generation unit first... the left-eye image I L (x, y) and the right-eye image IR Obtain (x, y) (step S1301). Next, I L (x, y) and I R Generate a disparity map D(x, y) from (x, y) and I (step S1302). For the generation of D(x, y), for example, an existing technique described in the following Reference 3 may be used. [Reference 3] A. Hosni et al., "Fast cost-volume filtering for visual correspondence and beyond," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 35, No. 2, pp.504-511, Feb. 2013.

[0028] <Generation of disparity remapping function> The procedure for generating the disparity remapping function according to the second embodiment is the same as that of the first embodiment. <Generation of multi-band disparity map> The procedure for generating the multi-band disparity map according to the second embodiment is the same as that of the first embodiment.

[0029] <Generation of disparity induction pattern> For the generation of the disparity induction pattern according to the second embodiment, instead of the two-dimensional image I(x, y) of the first embodiment, the left-eye image I L (x, y) is used. Other procedures are the same as those of the first embodiment. <Generation of Hidden Stereo image pair> The procedure for generating the Hidden Stereo image pair according to the second embodiment is the same as that of the first embodiment.

[0030] [Third Embodiment] In the second embodiment, a disparity map is generated from a stereo image pair using existing technology, but to obtain a disparity map with higher accuracy, a phase-based disparity correction process described in Reference 4 below may be performed. The correction process not only provides sub-pixel accuracy in disparity, but may also provide more robust results in situations where multiple disparities may exist for the same pixel, such as with glossy or semi-transparent objects. [Reference 4] P. Kellnhofer et al., "3DTV at home: eulerian-lagrangian stereo-to-multiview conversion," ACM Transaction on Graphics, vol. 36, No. 4, Article 146, pp.1-13, July 2017.

[0031] The functional block diagram of the stereoscopic image generating device according to the third embodiment is similar to the functional block diagram 12 of the second embodiment. <Disparity map generation> The procedure for generating a disparity map according to the third embodiment is the same as that according to the second embodiment. <Generating a parallax remapping function> The procedure for generating the disparity remapping function according to the third embodiment is the same as that according to the first embodiment.

[0032] <Generating a multi-band disparity map> FIG. 14 is a flowchart illustrating the operation of the multiband disparity map generator according to the third embodiment. The multi-band disparity map generator first generates a stereo image pair I L (x,y), I R (x, y), a disparity map D(x, y), and a disparity remapping function g are obtained (step S1401). Next, for the disparity map D(x, y), the horizontal moving average is calculated for each spatial frequency band f within a neighborhood window range with the number of pixels corresponding to that wavelength. To reduce the amount of calculation, the moving average process is performed independently for each horizontal scanning line, and the results are combined to obtain G' D f (x, y) is obtained (step S1402).

[0033] I L and I R is divided into scan line units, I L (x) and I R (x) is subjected to the same processing as in step S1002 of the first embodiment, and a one-dimensional band-pass image B L f (x), B R f (x) is obtained (step S1403). L f (x) is B as it is f It may be used as (x) in the parallax induction pattern generation unit.

[0034] Multi-band disparity map G' D f Based on (x), for each spatial frequency band f, L f (x) and B R f Find the corresponding point of (x). B L f x in (x) L B corresponding to the th pixel R f Pixel x of (x) R As, x L -G' D f (x L ) is calculated (step S1404). Next, B L f (x L ) and B R f (x R ) is calculated (step S1405). Next, G' is calculated by the following equation 4. D f (x) is corrected and G'' D f (x) is obtained (step S1406). G'' D f (x)=G' D f (x)+Δφ / ω f(Formula 4) Finally, G'' D f (x) is concatenated in the y direction to form G'' D f (x, y) is generated, and the disparity of each pixel is converted through the disparity remapping function g to obtain G D f (x, y) is obtained (step S1407).

[0035] <Generation of Disparity Inducing Pattern> In generating the disparity inducing pattern according to the third embodiment, instead of the 2D image I(x, y) of the first embodiment, the left-eye image I L (x, y) is used. Other procedures are the same as those of the first embodiment. <Generation of Hidden Stereo Image Pair> The procedure for generating the Hidden Stereo image pair according to the third embodiment is the same as that of the first embodiment.

[0036] [Program, Recording Medium] The various processes described above can be implemented by causing the recording unit 2020 of the computer 2000 shown in FIG. 15 to read a program for executing each step of the above method and causing operations in the control unit 2010, input unit 2030, output unit 2040, display unit 2050, etc.

[0037] The program describing this processing content can be recorded on a computer-readable recording medium. As the computer-readable recording medium, for example, any of a magnetic recording device, optical disk, magneto-optical recording medium, semiconductor memory, etc. may be used.

[0038] Also, the distribution of this program can be performed, for example, by selling, transferring, lending, etc. a portable recording medium such as a DVD or CD-ROM on which the program is recorded. Further, the program can be stored in the storage device of a server computer and transferred from the server computer to other computers via a network, thereby constituting a configuration for distributing this program.

[0039] A computer that executes such a program may first temporarily store the program recorded on a portable recording medium or transferred from a server computer in its own storage device. Then, when executing a process, the computer reads the program stored on its own recording medium and executes the process in accordance with the read program. Alternatively, the computer may read the program directly from a portable recording medium and execute the process in accordance with the program. Furthermore, the computer may execute the process in accordance with the received program each time a program is transferred from a server computer to the computer. Alternatively, the server computer may not transfer the program to the computer, but may execute the process through a so-called ASP (Application Service Provider) service, which realizes the processing function by issuing an execution instruction and obtaining the results. In this embodiment, the program includes information used for processing by a computer that is equivalent to a program (such as data that is not a direct instruction to the computer but has properties that define computer processing).

[0040] Furthermore, in this embodiment, the device is configured by executing a predetermined program on a computer, but at least a part of the processing contents may be realized by hardware.

Claims

1. A stereo image generating device that generates a stereo image from image information, a parallax map of the image, and viewer's viewpoint information, a parallax remapping function generating unit that extracts, from the observer's viewpoint information, a viewpoint-neighboring parallax amount, which is the parallax amount of pixels that belong to a certain range in the parallax map, and generates a parallax remapping function that sets a most frequent value of the viewpoint-neighboring parallax amount to zero and converts the parallax amount of the entire parallax map into a parallax amount within a predetermined range; a multiband disparity map generator that generates a first multiband disparity map by low-pass filtering the disparity map, and generates a second multiband disparity map by correcting the first multiband disparity map with the disparity remapping function; a parallax inducing pattern generating unit that performs band-pass decomposition on image information to generate a plurality of first band-pass images whose spatial frequency bands correspond to the respective maps of the first multi-band parallax map, generates a plurality of second band-pass images by shifting each of the plurality of first band-pass images by π / 2 in spatial phase, and generates a parallax inducing pattern based on an image obtained by weighting the second band-pass images by the values ​​of the second multi-band parallax map for each of the spatial frequency bands; an image pair generator that generates a stereo image pair by adding and subtracting the image information and the parallax inducing pattern; A stereo image generating device comprising:

2. Input stereo image pair I L and I R and viewpoint information of an observer, Above I L and the aforementioned I R a disparity map generator for generating a disparity map from the a parallax remapping function generating unit that extracts, from the viewpoint information of the observer, a viewpoint-neighboring parallax amount, which is a parallax amount of pixels that belong to a certain range in the parallax map, and generates a parallax remapping function that sets a most frequent value of the viewpoint-neighboring parallax amount to zero and converts the parallax amount of the entire parallax map into a parallax amount within a predetermined range; low-pass filtering the disparity map to generate a first multi-band disparity map; L and the aforementioned I R is band-pass decomposed to generate a plurality of band-pass images B whose spatial frequency bands correspond to the respective maps of the first multi-band disparity map. Ｌ f and multiple bandpass images B Ｒ ｆ and for each of the spatial frequency bands, Ｌ ｆ and the above B Ｒ ｆ and a multiband disparity map generator that corrects the first multiband disparity map based on the disparity remapping function, and generates a second multiband disparity map by correcting the corrected first multiband disparity map using the disparity remapping function. Above I L a parallax inducing pattern generating unit that generates a plurality of first band-pass images by band-pass decomposing the image, the spatial frequency bands of which correspond to the respective maps of the first multi-band parallax map, generates a plurality of second band-pass images by shifting the spatial phase of each of the plurality of first band-pass images by π / 2, and weights the second band-pass images by the values ​​of the second multi-band parallax map for each of the spatial frequency bands, and generates a parallax inducing pattern based on the images obtained; Above I L an image pair generator that generates a stereo image pair by adding and subtracting the parallax inducing pattern; A stereo image generating device comprising:

3. 2. The stereo image generating device according to claim 1, the multiband disparity map generator processes the disparity map for each horizontal scan line; The parallax-inducing pattern generator processes the image for each horizontal scanning line. Stereo image generator.

4. 1. A method for generating a stereo image from image information, a disparity map of the image, and observer viewpoint information, comprising: a disparity remapping function generating unit extracting, from the viewpoint information of the observer, a viewpoint-neighboring disparity amount which is a disparity amount of pixels belonging to a certain range in the disparity map, and generating a disparity remapping function which sets a most frequent value of the viewpoint-neighboring disparity amount to zero and converts the disparity amount of the entire disparity map into a disparity amount within a predetermined range; a multiband disparity map generator low-pass filtering the disparity map to generate a first multiband disparity map, and correcting the first multiband disparity map with the disparity remapping function to generate a second multiband disparity map; a parallax inducing pattern generating unit performing band-pass decomposition on image information to generate a plurality of first band-pass images whose spatial frequency bands correspond to the respective maps of the first multi-band parallax map, generating a plurality of second band-pass images by shifting each of the plurality of first band-pass images by π / 2 in spatial phase, and generating a parallax inducing pattern based on images obtained by weighting the second band-pass images by the values ​​of the second multi-band parallax map for each of the spatial frequency bands; an image pair generating unit adding and subtracting the image information and the parallax inducing pattern to generate a stereo image pair; A stereo image generation method including:

5. A program for causing a computer to function as the stereo image generating device according to any one of claims 1 to 3.

Images