Sony Patent | Image Processing Apparatus And Image Processing Method

Patent: Image Processing Apparatus And Image Processing Method

Publication Number: 20200051323

Publication Date: 20200213

Applicants: Sony

Abstract

The present disclosure relates to an image processing apparatus and an image processing method that enable generation of only a 3D model of a foreground. A reconstruction unit generates a 3D model of the foreground, on the basis of depth images of a plurality of viewpoints and foreground images of the plurality of viewpoints. The present disclosure can be applied to a decoding apparatus or the like that decodes an encoded stream of depth-related images and color images of a 3D model of a plurality of viewpoints, and generates a 3D model of the foreground on the basis of the resultant depth-related images and color images, and virtual viewpoint information including internal parameters and external parameters for virtual cameras of the respective viewpoints, for example.

TECHNICAL FIELD

[0001] The present disclosure relates to an image processing apparatus and an image processing method, and more particularly, to an image processing apparatus and an image processing method that enable generation of only a 3D model of a foreground.

BACKGROUND ART

[0002] There is a technique for generating a 3D model of an object from color images and depth images captured by a plurality of cameras (see Non-Patent Document 1, for example).

CITATION LIST

Non-Patent Document

[0003] Non-Patent Document 1: Saied Moezzi, Li-Cheng Tai, and Philippe Gerard, “Virtual View Generation for 3D Digital Video”, University of California,* San Diego*

SUMMARY OF THE INVENTION

Problems to be Solved by the Invention

[0004] However, any method has not been devised to generate only a 3D model of a foreground.

[0005] The present disclosure is made in view of such circumstances, and is to enable generation of only a 3D model of the foreground.

Solutions to Problems

[0006] An image processing apparatus of a first aspect of the present disclosure is an image processing apparatus that includes a 3D model generation unit that generates a 3D model of the foreground, on the basis of depth images of a plurality of viewpoints and foreground images of the plurality of viewpoints.

[0007] An image processing method of the first aspect of the present disclosure is compatible with the image processing apparatus of the first aspect of the present disclosure.

[0008] In the first aspect of the present disclosure, a 3D model of the foreground is generated, on the basis of depth images of a plurality of viewpoints and the foreground images of the plurality of viewpoints.

[0009] An image processing apparatus of a second aspect of the present disclosure is an image processing apparatus that includes a transmission unit that transmits depth images of a plurality of viewpoints and foreground information about the foreground of the plurality of viewpoints.

[0010] An image processing method of the second aspect of the present disclosure is compatible with the image processing apparatus of the second aspect of the present disclosure.

[0011] In the second aspect of the present disclosure, depth images of a plurality of viewpoints and foreground information about the foreground of the plurality of viewpoints are transmitted.

[0012] An image processing apparatus of a third aspect of the present disclosure is an image processing apparatus that includes a 3D model generation unit that generates a 3D model of the foreground, on the basis of foreground depth images of a plurality of viewpoints.

[0013] An image processing method of the third aspect of the present disclosure is compatible with the image processing apparatus of the third aspect of the present disclosure.

[0014] In the third aspect of the present disclosure, a 3D model of the foreground is generated, on the basis of foreground depth images of a plurality of viewpoints.

[0015] Note that the image processing apparatuses of the first through third aspects can also be formed by a computer executing a program.

[0016] Further, to obtain the image processing apparatuses of the first through third aspects, the program to be executed by the computer may be transmitted and provided via a transmission medium, or the program recorded on a recording medium may be provided.

Effects of the Invention

[0017] According to the first and third aspects of the present disclosure, only a 3D model of a foreground can be generated.

[0018] Further, according to the second aspect of the present disclosure, it is possible to transmit information that enables generation of only a 3D model of a foreground.

[0019] Note that effects of the present technology are not limited to the effects described above, and may include any of the effects described in the present disclosure.

BRIEF DESCRIPTION OF DRAWINGS

[0020] FIG. 1 is a block diagram showing an example configuration of a first embodiment of an image processing system to which the present disclosure is applied.

[0021] FIG. 2 is a graph showing the relationship between a distance Z and a pixel value p.

[0022] FIG. 3 is a flowchart for explaining an encoding process to be performed by the encoding apparatus shown in FIG. 1.

[0023] FIG. 4 is a block diagram showing an example configuration of the reconstruction unit shown in FIG. 1.

[0024] FIG. 5 is a diagram for explaining a method of generating a 3D model of a foreground with the reconstruction unit shown in FIG. 4.

[0025] FIG. 6 is a diagram for explaining a method of generating a 3D model of a foreground with the reconstruction unit shown in FIG. 4.

[0026] FIG. 7 is a diagram for explaining a method of generating a 3D model of a foreground with the reconstruction unit shown in FIG. 4.

[0027] FIG. 8 is a diagram for explaining a method of generating a 3D model of a foreground with the reconstruction unit shown in FIG. 4.

[0028] FIG. 9 is a diagram for explaining a method of generating a 3D model of a foreground with the reconstruction unit shown in FIG. 4.

[0029] FIG. 10 is a flowchart for explaining a decoding process to be performed by the decoding apparatus shown in FIG. 1.

[0030] FIG. 11 is a block diagram showing the configuration of a reconstruction unit according to a second embodiment of an image processing system to which the present disclosure is applied.

[0031] FIG. 12 is a block diagram showing the configuration of a reconstruction unit according to a third embodiment of an image processing system to which the present disclosure is applied.

[0032] FIG. 13 is a diagram showing an example of a convex hull.

[0033] FIG. 14 is a block diagram showing an example configuration of a reconstruction unit according to a fourth embodiment of an image processing system to which the present disclosure is applied.

[0034] FIG. 15 is a diagram showing example depth images of foregrounds.

[0035] FIG. 16 is a diagram for explaining a method of generating a polygon mesh.

[0036] FIG. 17 is a diagram showing examples of polygon meshes.

[0037] FIG. 18 is a diagram for explaining a method of superimposing polygon meshes of viewpoints of respective virtual cameras.

[0038] FIG. 19 is a flowchart for explaining a decoding process to be performed by a decoding apparatus according to the fourth embodiment.

[0039] FIG. 20 is a block diagram showing the configuration of a reconstruction unit according to a fifth embodiment of an image processing system to which the present disclosure is applied.

[0040] FIG. 21 is a block diagram showing the configuration of a reconstruction unit according to a sixth embodiment of an image processing system to which the present disclosure is applied.

[0041] FIG. 22 is a block diagram showing an example configuration of the hardware of a computer.

[0042] FIG. 23 is a block diagram schematically showing an example configuration of a vehicle control system.

[0043] FIG. 24 is an explanatory diagram showing an example of installation positions of an external information detector and imaging units.

MODES FOR CARRYING OUT THE INVENTION

[0044] The following is a description of modes (hereinafter referred to as embodiments) for carrying out the present disclosure. Note that explanation will be made in the following order.

[0045] 1. First embodiment: Image processing system (FIGS. 1 through 10)

[0046] 2. Second embodiment: Image processing system (FIG. 11)

[0047] 3. Third embodiment Image processing system (FIGS. 12 and 13)

[0048] 4. Fourth embodiment: Image processing system (FIGS. 14 through 19)

[0049] 5. Fifth embodiment: Image processing system (FIG. 20)

[0050] 6. Sixth embodiment: Image processing system (FIG. 21)

[0051] 7. Seventh embodiment: Computer (FIG. 22)

[0052] 8. Example Applications (FIGS. 23 and 24)

FIRST EMBODIMENT

Example Configuration of an Image Processing System

[0053] FIG. 1 is a block diagram showing an example configuration of a first embodiment of an image processing system to which the present disclosure is applied.

[0054] An image processing system 10 in FIG. 1 includes an imaging apparatus 11, an encoding apparatus 12 (image processing apparatus), a decoding apparatus 13 (image processing apparatus), and a display device 14. The image processing system 10 generates and displays a color image of a display viewpoint, using a color image and a depth image acquired by the imaging apparatus 11.

[0055] Specifically, the imaging apparatus 11 of the image processing system 10 includes a multi-view camera, a distance measuring instrument, and an image processing unit, for example. The multi-view camera of the imaging apparatus 11 is formed with a plurality of cameras. The respective cameras capture moving images of color images of an object, the moving images having at least one common portion. The distance measuring instrument is provided in each camera, for example, and generates a moving image of a depth image having the same viewpoint as that camera.

[0056] The image processing unit of the imaging apparatus 11 generates a 3D model of the object by calculating a visual hull or the like for each frame, using the moving images of the color images and the depth images of the viewpoints of the respective cameras, and external parameters and internal parameters of the respective cameras. The image processing unit generates the 3D data of the object, which is shape information (connectivity) indicating the three-dimensional positions of the vertices of the respective polygon meshes constituting the 3D model and the connection between the vertices, and color information about the polygon meshes.

[0057] The method adopted for generating the 3D data at the image processing unit may be the method described in Non-Patent Document 1 or the like, for example. Note that the 3D data may contain shape information and color images of the viewpoints of the respective cameras. The image processing unit supplies the 3D data to the encoding apparatus 12.

[0058] The encoding apparatus 12 includes a conversion unit 21, a generation unit 22, an encoding unit 23, a storage unit 24, and a transmission unit 25.

[0059] The conversion unit 21 of the encoding apparatus 12 determines a plurality of viewpoints of a color image and a depth image of the 3D model to be generated. Here, it is assumed that the viewpoints of the color image and the depth image to be generated are the same. However, the viewpoints and the number of viewpoints of the color image and the depth image may differ.

[0060] The conversion unit 21 generates external parameters and internal parameters for virtual cameras of the plurality of viewpoints that have been determined. On the basis of the external parameters and the internal parameters for the respective virtual cameras, the conversion unit 21 generates, from the 3D data supplied on a frame-by-frame basis from the imaging apparatus 11, a color image of each frame acquired by each virtual camera and a depth image corresponding to the color image.

[0061] The method adopted for generating a color image and a depth image from the 3D data may be the method disclosed by Masayuki Tanimoto in “Realizing the Ultimate Visual Communication”, IEICE Technical Report, CS, Communication Systems vol. 110 (no. 323), pp. 73-78, Nov. 25, 2010, and the like, for example.

[0062] For example, the depth image may be an image that has a pixel value obtained by quantizing the distance Z in the depth direction between the viewpoint and the object at each pixel. In this case, the pixel value p of each pixel in the depth image is expressed by the following expression (1), for example.

[Expression 1]

p=(Z-Zmin)/(Zmax-Zmin).times.((1<<bitdepth)-1) (1)

[0063] Note that Zmin and Zmax represent the minimum value and the maximum value of the distance Z, respectively. Further, “bitdepth” represents the bit width of the pixel value p. According to the expression (1), the pixel value p is a value obtained by quantizing the distance Z in the range from the minimum value Zmin to the maximum value Zmax into a bit with the bit width “bitdepth”. The greater the pixel value p, the longer the distance Z (the object is farther from the viewpoint). The smaller the pixel value p, the shorter the distance Z (the object is closer to the viewpoint). The minimum value Zmin and the maximum value Zmax may vary with each viewpoint, or may be the same for all the viewpoints.

[0064] Further, the depth image may be an image that has a pixel value obtained by quantizing the reciprocal 1/Z of the distance Z at each pixel. In this case, the pixel value p of each pixel in the depth image is expressed by the following expression (2).

[Expression 2]

p=(1/Z-1/Zmax)/(1/Zmin-1/Zmax).times.((1<<bitdepth)-1) (2)

[0065] According to the expression (2), the pixel value p is a value obtained by quantizing the reciprocal 1/Z in the range from the minimum value 1/Zmax to the maximum value 1/Zmin into a bit with the bit width “bitdepth”. The smaller the pixel value p, the longer the distance Z (the object is farther from the viewpoint). The greater the pixel value p, the shorter the distance Z (the object is closer to the viewpoint).

[0066] Note that the calculation formula for determining the pixel value p may be other than the expressions (1) and (2). The calculation formula for determining the pixel value p may vary with each viewpoint, or may be the same for all the viewpoints.

[0067] The conversion unit 21 supplies the color image of each virtual camera to the generation unit 22 and the encoding unit 23, and supplies the depth image to the encoding unit 23. The conversion unit 21 also supplies the storage unit 24 with the external parameters and the internal parameters for the respective virtual cameras as virtual viewpoint information.

[0068] For each virtual camera, the generation unit 22 generates a silhouette image showing a foreground silhouette as foreground information about the foreground of the viewpoint of the virtual camera, from the color image supplied from the conversion unit 21. Specifically, for each virtual camera, the generation unit 22 extracts a color image of the background from the color image of the entire 3D model supplied from the conversion unit 21. The generation unit 22 then generates a difference between the color image of the entire 3D model and the color image of the background as a silhouette image for each virtual camera. As a result, the silhouette image becomes an image that is white (the pixel value being 255) only in the foreground region on which the 3D model of the foreground in the color image of the entire 3D model of each virtual camera is projected, and is black (the pixel value being 0) in the background region.

[0069] Note that the pixel value of the foreground region of the silhouette image may be the ID assigned to the 3D model of the foreground corresponding to the foreground region. The generation unit 22 generates a depth-related image of YUV 420 having the depth image supplied from the conversion unit 21 as the luminance component and the silhouette image as the color component, and supplies the depth-related image to the encoding unit 23.

[0070] The encoding unit 23 encodes the color image of each virtual camera supplied from the conversion unit 21, and the depth-related image of each virtual camera supplied from the generation unit 22. The encoding method adopted herein may be Advanced Video Coding (AVC), High Efficiency Video Coding (HEVC), or the like. The encoding unit 23 supplies the encoded stream obtained as a result of the encoding to the storage unit 24.

[0071] The storage unit 24 stores the virtual viewpoint information supplied from the conversion unit 21, and the encoded stream supplied from the encoding unit 23.

[0072] The transmission unit 25 reads the virtual viewpoint information as metadata from the storage unit 24, and transmits the metadata to the decoding apparatus 13. The transmission unit 25 also reads the encoded stream, and transmits the encoded stream to the decoding apparatus 13.

[0073] As described above, the encoding apparatus 12 encodes a depth image and a silhouette image as one depth-related image, and transmits the depth-related image to the decoding apparatus 13. Therefore, the network bandwidth between the encoding apparatus 12 and the decoding apparatus 13 can be made smaller than in a case where the depth image and the silhouette image are encoded as separate images and are transmitted as separate images to the decoding apparatus 13

[0074] The decoding apparatus 13 includes a reception unit 31, a decoding unit 32, a reconstruction unit 33, and a rendering unit 34. External parameters and internal parameters for a virtual camera of the display viewpoint are input from the user viewing the display device 14 to the decoding apparatus 13, and are supplied as display viewpoint information to the rendering unit 34.

[0075] The reception unit 31 of the decoding apparatus 13 receives the virtual viewpoint information and the encoded stream transmitted from the transmission unit 25 of the encoding apparatus 12. The reception unit 31 supplies the virtual viewpoint information to the reconstruction unit 33, and supplies the encoded stream to the decoding unit 32.

[0076] The decoding unit 32 decodes the encoded stream supplied from the reception unit 31 by a method compatible with the encoding method at the encoding unit 23. The decoding unit 32 supplies the resultant color image and depth-related image of each virtual camera to the reconstruction unit 33.

[0077] On the basis of the virtual viewpoint information supplied from the reception unit 31 and the color images and the depth-related images supplied from the decoding unit 32, the reconstruction unit 33 (the 3D model generation unit) reconstructs (generates) only the 3D data of the 3D model of the foreground.

[0078] Note that the number of viewpoints of the depth-related images to be used for reconstruction of the foreground 3D data may be any number that is equal to or smaller than the number of viewpoints of the depth-related images transmitted from the encoding apparatus 12. The larger the number of viewpoints corresponding to the depth-related images to be used for reconstruction of the 3D model of the foreground, the higher the accuracy of the 3D model of the foreground. The depth-related images to be used for reconstruction of the 3D data of the foreground may be determined in accordance with the capability of the decoding apparatus 13 and the state of the network bandwidth between the encoding apparatus 12 and the decoding apparatus 13. The reconstruction unit 33 supplies the 3D data of the foreground to the rendering unit 34.

[0079] Like the conversion unit 21, on the basis of the display viewpoint information, the rendering unit 34 generates a foreground color image of the display viewpoint as a display image from the 3D data supplied from the reconstruction unit 33, and supplies the foreground display image to the display device 14.

[0080] The display device 14 is formed with a two-dimensional head mounted display (HMD), a two-dimensional monitor, or the like. The display device 14 two-dimensionally displays a display image on the basis of the display image supplied from the reconstruction unit 33.

[0081] Note that the display device 14 may be formed with a three-dimensional head mounted display, a three-dimensional monitor, or the like. In this case, the rendering unit 34 generates a foreground depth image of the display viewpoint from the 3D data on the basis of the display viewpoint information like the conversion unit 21, and supplies the foreground depth image to the display device 14. The display device 14 three-dimensionally displays a display image on the basis of the display image and the depth image supplied from the rendering unit 34.

[0082] Further, in the example shown in FIG. 1, the 3D model is generated through calculation of a visual hull or the like, but may be generated with point clouds. In this case, the 3D data includes the position and color information about each point cloud, or the position of each point cloud and a color image of the viewpoint of each camera.

[0083] As described above, in the image processing system 10, the encoding apparatus 12 performs encoding by converting the 3D data into color images and depth images of virtual cameras of a plurality of viewpoints. Accordingly, the encoding method that can be adopted here may be a highly-compressed two-dimensional moving image encoding method such as AVC or HEVC. As a result, information indicating a 3D model can be transmitted at a lower bit rate than in a case where 3D data is transmitted as it is.

Relationship between the Distance Z and the Pixel Value p

[0084] FIG. 2 is a graph showing the relationship between the distance Z and the pixel value p obtained according to the expressions (1) and (2).

[0085] In the graph in FIG. 2, the abscissa axis indicates the pixel value p, and the ordinate axis indicates the distance Z. Further, in the example shown in FIG. 2, the minimum value Zmin is 1000 mm, the maximum value Zmax is 10000 mm, and the bit width “bitdepth” is 5 bits.

[0086] In this case, according to the expression (1), the distance Z range of 1000 mm to 10000 mm is divided into 32 equal portions, and different pixel values p are assigned to the respective portions of the divided distance Z range, as indicated by a dotted line in FIG. 2. Therefore, the distance Z range corresponding to each pixel value p is the same. In other words, the quantization step is constant.

[0087] On the other hand, according to the expression (2), the range of the reciprocal 1/Z of the distance Z, which is from 1/10000 mm to 1/1000 mm, is divided into 32 equal portions, and different pixel values p are assigned to the respective portions of the divided reciprocal 1/Z range, as indicated by a solid line in FIG. 2. Accordingly, the smaller the pixel value p, the wider the distance Z range corresponding to that value. The greater the pixel value p, the narrower the distance Z range corresponding to that value. In other words, where the object is farther from the viewpoint, the quantization step of the pixel value p is larger. Where the object is closer to the viewpoint, the quantization step of the pixel value p is smaller. Since the error of the distance Z affects the accuracy of a 3D model greater when the object is located closer to the viewpoint, the accuracy of the 3D model can be increased by determining the pixel value according to the expression (2).

[0088] Note that, in the description below, the pixel value of a depth image is determined according to the expression (1), unless otherwise specified.

Description of an Encoding Process at the Encoding Apparatus

[0089] FIG. 3 is a flowchart for explaining an encoding process to be performed by the encoding apparatus 12 shown in FIG. 1. This encoding process is started when 3D data is supplied from the imaging apparatus 11 on a frame-by-frame basis, for example.

[0090] In step S11 in FIG. 3, the conversion unit 21 of the encoding apparatus 12 determines a plurality of viewpoints of a color image and a depth image of the 3D model to be generated.

[0091] In step S12, the conversion unit 21 generates external parameters and internal parameters for virtual cameras of the determined plurality of viewpoints as virtual viewpoint information, and supplies the virtual viewpoint information to the storage unit 24.

[0092] In step S13, on the basis of the virtual viewpoint information, the conversion unit 21 generates color images and depth images of the viewpoints of the respective virtual cameras from the 3D data supplied on a frame-by-frame basis from the imaging apparatus 11. The conversion unit 21 supplies the color images of the viewpoints of the respective virtual cameras to the generation unit 22 and the encoding unit 23, and supplies the depth images to the generation unit 22.

[0093] In step S14, the generation unit 22 generates a silhouette image from the color images supplied from the conversion unit 21 for the respective virtual cameras.

[0094] In step S15, the generation unit 22 generates a depth-related image for each virtual camera, using the depth image supplied from the conversion unit 21 as the luminance component and the silhouette image as the color component. The generation unit 22 then supplies the depth-related image to the encoding unit 23.

[0095] In step S16, the encoding unit 23 encodes the color image of each virtual camera supplied from the conversion unit 21, and the depth-related image of each virtual camera supplied from the generation unit 22. The encoding unit 23 supplies the encoded stream obtained as a result of the encoding to the storage unit 24.

[0096] In step S17, the storage unit 24 stores the virtual viewpoint information supplied from the conversion unit 21, and the encoded stream supplied from the encoding unit 23.

[0097] In step S18, The transmission unit 25 reads the virtual viewpoint information and the encoded stream stored in the storage unit 24, and transmits the virtual viewpoint information and the encoded stream to the decoding apparatus 13.

Example Configuration of the Reconstruction Unit

[0098] FIG. 4 is a block diagram showing an example configuration of the reconstruction unit 33 shown in FIG. 1.

[0099] The reconstruction unit 33 in FIG. 4 includes a visual hull generation unit 101, a correction unit 102, a mesh generation unit 103, and a 3D data generation unit 104.

[0100] The visual hull generation unit 101 of the reconstruction unit 33 generates a visual hull, on the basis of the virtual viewpoint information supplied from the reception unit 31 in FIG. 1, and the silhouette image as the color component of the depth-related image of each viewpoint supplied from the decoding unit 32. A visual hull is the intersection space of a cone that is formed, for each camera, from the optical centers of a plurality of cameras and the silhouettes of the object captured by the cameras. The visual hull generation unit 101 supplies the generated visual hull to the correction unit 102.

[0101] The correction unit 102 corrects the visual hull supplied from the visual hull generation unit 101, on the basis of the depth images as the luminance components of the depth-related images of the respective viewpoints supplied from the decoding unit 32. By doing so, the correction unit 102 generates a 3D model of the foreground. The correction unit 102 supplies the 3D model of the foreground to the mesh generation unit 103.

[0102] The mesh generation unit 103 converts the 3D model (Voxel) of the foreground into one or more polygon meshes. The mesh generation unit 103 supplies shape information about the respective polygon meshes of the 3D model of the foreground to the 3D data generation unit 104.

[0103] On the basis of the virtual viewpoint information supplied from the reception unit 31 and the color images of the respective viewpoints supplied from the decoding unit 32, the 3D data generation unit 104 generates color information about the polygon meshes corresponding to the respective pieces of the shape information supplied from the mesh generation unit 103. The 3D data generation unit 104 supplies the shape information and the color information about the respective polygon meshes as the 3D data of the 3D model of the foreground to the rendering unit 34 in FIG. 1.

Description of a Method of Generating a 3D Model of the Foreground

[0104] FIGS. 5 through 9 are diagrams for explaining a method of generating a 3D model of the foreground with the reconstruction unit 33 shown in FIG. 4.

[0105] In the example shown in FIGS. 5 through 9, the shape of a foreground object 121 is a triangular prism. FIGS. 5 through 9 are also views of the object 121 as viewed from above the virtual cameras.

[0106] Further, in the example shown in FIGS. 5 through 9, the viewpoints of the virtual cameras are a total of four viewpoints A through D that are arranged to surround the object 121, as shown in FIG. 5. Further, the screen (projection plane) 131 of the viewpoint A is located within an angle of view 141 around the viewpoint A. Like the screen 131 of the viewpoint A, the screens 132 through 134 of the viewpoints B through D are also located within an angle of view 142, an angle of view 143, and an angle of view 144, respectively.

[0107] In this case, the silhouette images of viewpoints A through D in FIG. 5 are silhouette images 151 through 154 shown in FIG. 6. Accordingly, the visual hull generation unit 101 generates a visual hull 170 shown in FIG. 7, on the basis of virtual viewpoint information about the viewpoints A through D and the silhouette images 151 through 154.

[0108] Meanwhile, the depth images of the viewpoints A through D in FIG. 5 are depth images 191 through 194 shown in FIG. 8. Therefore, as shown in A of FIG. 9, the correction unit 102 first corrects the visual hull 170 generated by the visual hull generation unit 101, on the basis of the depth image 191. By doing so, the correction unit 102 generates a visual hull 201. Specifically, the correction unit 102 recognizes, from the depth image 191, that the distance Z in the depth direction between the entire viewpoint-A side surface of the 3D model of the foreground to be generated and the viewpoint A is constant. However, the distance Z in the depth direction between the entire viewpoint-A side surface of the visual hull 170 and the viewpoint A is not constant. Therefore, the correction unit 102 deletes a convex portion 170A of the viewpoint-A side surface of the visual hull 170 so that the distance Z becomes constant. By doing so, the correction unit 102 generates the visual hull 201.

[0109] The correction unit 102 then corrects the visual hull 201 on the basis of the depth image 192, and generates a visual hull 202, as shown in B of FIG. 9. Specifically, from the depth image 192, the correction unit 102 recognizes that the distance Z in the depth direction between the viewpoint-B side surface of the 3D model of the foreground to be generated and the viewpoint B increases in the direction toward the right as viewed from the viewpoint B. However, the distance Z in the depth direction between the viewpoint-B side surface of the visual hull 202 and the viewpoint B increases toward the right as viewed from the viewpoint B, but does not change significantly. Therefore, the correction unit 102 deletes a convex portion 201A of the viewpoint-B side surface of the visual hull 201 so that the distance Z increases in the direction toward the right as viewed from the viewpoint B. By doing so, the correction unit 102 generates the visual hull 202.

[0110] The correction unit 102 then corrects the visual hull 202 on the basis of the depth image 193, and generates a visual hull 203, as shown in C of FIG. 9. Specifically, from the depth image 193, the correction unit 102 recognizes that the distance Z in the depth direction between the viewpoint-C side surface of the 3D model of the foreground to be generated and the viewpoint C increases in the direction toward the left or the right from a predetermined position, as viewed from the viewpoint C.

[0111] On the left side of the predetermined position on the viewpoint-C side surface of the visual hull 202 as viewed from the viewpoint C, the distance Z in the depth direction from the viewpoint C increases in the direction toward the left. Therefore, the correction unit 102 does not perform any correction on the left side. However, on the right side of the predetermined position on the viewpoint-C side surface of the visual hull 202 as viewed from the viewpoint C, the distance in the depth direction from the viewpoint C increases in the direction toward the right but does not change significantly. Therefore, the correction unit 102 deletes a convex portion 202A of the visual hull 202 on the right side of the predetermined position as viewed from the viewpoint C, so that the distance Z increases in the direction toward the right from the predetermined position as viewed from the viewpoint C. By doing so, the correction unit 102 generates the visual hull 203.

[0112] Finally, the correction unit 102 generates the visual hull 203 as a corrected visual hull 203 on the basis of the depth image 194, as shown in D of FIG. 9. Specifically, from the depth image 194, the correction unit 102 recognizes that the distance Z in the depth direction between the viewpoint-D side surface of the 3D model of the foreground to be generated and the viewpoint D increases in the direction toward the left as viewed from the viewpoint D. Since the distance Z in the depth direction between the viewpoint-D side surface of the visual hull 203 and the viewpoint D increases in the direction toward the left as viewed from the viewpoint D, the correction unit 102 does not perform any correction on the visual hull 203.

[0113] As described above, the correction unit 102 corrects the visual hull 170 on the basis of the depth images 191 through 194, to generate the visual hull 203 having the same shape as the foreground object 121. The correction unit 102 then supplies the visual hull 203 as the 3D model of the foreground to the mesh generation unit 103.

Description of a Process at the Decoding Apparatus

[0114] FIG. 10 is a flowchart for explaining a decoding process to be performed by the decoding apparatus 13 shown in FIG. 1. This decoding process is started when the encoded stream and the virtual viewpoint information are transmitted frame by frame from the transmission unit 25 of the encoding apparatus 12, for example.

[0115] In step S31 in FIG. 10, the reception unit 31 of the decoding apparatus 13 receives the virtual viewpoint information and the encoded stream transmitted from the transmission unit 25 of the encoding apparatus 12. The reception unit 31 supplies the virtual viewpoint information to the reconstruction unit 33, and supplies the encoded stream to the decoding unit 32.

[0116] In step S32, the decoding unit 32 decodes the encoded stream supplied from the reception unit 31 by a method compatible with the encoding method at the encoding unit 23. The decoding unit 32 supplies the resultant color image and depth-related image of each virtual camera to the reconstruction unit 33.

[0117] In step S33, the visual hull generation unit 101 (FIG. 4) of the reconstruction unit 33 generates a visual hull, on the basis of the virtual viewpoint information supplied from the reception unit 31, and the silhouette image as the color component of the depth-related image of each virtual camera supplied from the decoding unit 32. The visual hull generation unit 101 supplies the generated visual hull to the correction unit 102.

[0118] In step S34, the correction unit 102 corrects the visual hull supplied from the visual hull generation unit 101, on the basis of the depth images as the luminance components of the depth-related images of the respective virtual cameras supplied from the decoding unit 32. By doing so, the correction unit 102 generates a 3D model of the foreground. The correction unit 102 supplies the 3D model of the foreground to the mesh generation unit 103.

[0119] In step S35, the mesh generation unit 103 converts the 3D model of the foreground into one or more polygon meshes. The mesh generation unit 103 supplies shape information about the respective polygon meshes of the 3D model of the foreground to the 3D data generation unit 104.

[0120] In step S36, the 3D data generation unit 104 reconstructs the 3D data of the 3D model of the foreground, on the basis of the virtual viewpoint information, the color images of the respective virtual cameras, and the shape information about the respective polygon meshes. The 3D data generation unit 104 supplies the 3D data of the 3D model of the foreground to the rendering unit 34.

[0121] In step S37, like the conversion unit 21, the rendering unit 34 generates a foreground color image of the display viewpoint as a display image from the 3D data of the 3D model of the foreground supplied from the reconstruction unit 33, on the basis of the display viewpoint information. The rendering unit 34 supplies the foreground display image to the display device 14.

[0122] As described above, the encoding apparatus 12 transmits silhouette images of a plurality of viewpoints as foreground information, together with depth images of the plurality of viewpoints, to the decoding apparatus 13. Thus, the decoding apparatus 13 can generate only a 3D model of the foreground, on the basis of the depth images and the silhouette images of the plurality of viewpoints.

SECOND EMBODIMENT

Example Configuration of the Reconstruction Unit

[0123] A second embodiment of an image processing system to which the present disclosure is applied differs from the first embodiment in that the foreground information is not silhouette images of the respective virtual cameras but thresholds for the pixel value of the foreground in depth images of the respective virtual cameras (the thresholds will be hereinafter referred to as the foreground depth thresholds).

[0124] Specifically, the configuration of the second embodiment of an image processing system to which the present disclosure is applied is similar to the configuration shown in FIG. 1, except that the generation unit 22 generates foreground depth images in place of silhouette images, depth-related images are replaced with depth images, metadata is replaced with virtual viewpoint information and the foreground depth thresholds, and the reconstruction unit 33 has a different configuration. Therefore, explanation of the components other than the reconstruction unit 33 will not be unnecessarily repeated below.

[0125] FIG. 11 is a block diagram showing the configuration of the reconstruction unit 33 according to the second embodiment of an image processing system to which the present disclosure is applied.

[0126] In the configuration shown in FIG. 11, the same components as those shown in FIG. 4 are denoted by the same reference numerals as those used in FIG. 4. The explanations that have already been made will not be repeated.

[0127] The configuration of the reconstruction unit 33 in FIG. 11 differs from the configuration in FIG. 4 in that a silhouette image generation unit 221 is added.

[0128] In the second embodiment, the reception unit 31 of the decoding apparatus 13 receives an encoded stream of color images and depth images of a plurality of virtual cameras, supplies the encoded stream to the decoding unit 32, receives the foreground depth thresholds and virtual viewpoint information, and supplies the foreground depth thresholds and the virtual viewpoint information to the reconstruction unit 33. The decoding unit 32 decodes the encoded stream, and supplies the resultant color images and depth images of the plurality of virtual cameras to the reconstruction unit 33.

[0129] The silhouette image generation unit 221 (the image generation unit) of the reconstruction unit 33 generates a silhouette image for each virtual camera, on the basis of the input foreground depth thresholds and depth images.

[0130] Specifically, for each pixel of the depth images, the silhouette image generation unit 221 determines whether or not the pixel value of the pixel is equal to or smaller than the foreground depth threshold. The silhouette image generation unit 221 sets the pixel value of a silhouette image of a pixel whose pixel value is determined to be equal to or smaller than the foreground depth threshold at 255, which represents a foreground region, and sets the pixel value of a silhouette image of a pixel whose pixel value is determined to be greater than the foreground depth threshold at 0, which represents a background region.

[0131] Note that, in a case where the pixel value of a depth image is determined according to the above expression (2), the pixel value of a silhouette image of a pixel whose pixel value is determined to be equal to or smaller than the foreground depth threshold is set at 0, and the pixel value of a silhouette image of a pixel whose pixel value is determined to be greater than the foreground depth threshold is set at 255.

[0132] The silhouette image generation unit 221 can generate a silhouette image as described above. The silhouette image generation unit 221 supplies the silhouette images to the visual hull generation unit 101.

[0133] Note that a decoding process in the second embodiment is similar to the decoding process shown in FIG. 10, except that a process in which the silhouette image generation unit 221 generates silhouette images is performed before the process in step S33.

[0134] As described above, in the second embodiment, the encoding apparatus 12 transmits foreground depth thresholds for a plurality of viewpoints as foreground information, together with depth images of the plurality of viewpoints, to the decoding apparatus 13. Thus, the decoding apparatus 13 can generate only a 3D model of the foreground, on the basis of the depth images and the foreground depth thresholds of the plurality of viewpoints.

THIRD EMBODIMENT

Example Configuration of the Reconstruction Unit

[0135] A third embodiment of an image processing system to which the present disclosure is applied differs from the first embodiment in that any foreground information is not transmitted, and a depth-related image is an image that has a depth image of the foreground as the luminance component, and a depth image of the background as the color component.

[0136] Specifically, the configuration of the third embodiment of an image processing system to which the present disclosure is applied is similar to the configuration shown in FIG. 1, except that the conversion unit 21 generates depth-related images by generating depth images of the foreground separately from depth images of the background, the generation unit 22 is not provided, and the reconstruction unit 33 has a different configuration. Therefore, explanation of the components other than the reconstruction unit 33 will not be unnecessarily repeated below.

[0137] FIG. 12 is a block diagram showing the configuration of the reconstruction unit 33 according to the third embodiment of an image processing system to which the present disclosure is applied.

[0138] In the configuration shown in FIG. 12, the same components as those shown in FIG. 4 are denoted by the same reference numerals as those used in FIG. 4. The explanations that have already been made will not be repeated.

[0139] The configuration of the reconstruction unit 33 shown in FIG. 12 differs from the configuration in FIG. 4 in that a convex hull generation unit 241 and a correction unit 242 are provided in place of the visual hull generation unit 101 and the correction unit 102.

[0140] In the third embodiment, the reception unit 31 of the decoding apparatus 13 receives an encoded stream of color images and depth-related images of a plurality of virtual cameras, supplies the encoded stream to the decoding unit 32, receives virtual viewpoint information, and supplies the virtual viewpoint information to the reconstruction unit 33. The decoding unit 32 decodes the encoded stream, and supplies the resultant color images and depth images of the plurality of virtual cameras to the reconstruction unit 33.

[0141] On the basis of the input virtual viewpoint information, the convex hull generation unit 241 of the reconstruction unit 33 generates a convex hull (convex polygon) included in the angles of view of all the virtual cameras, and supplies the convex hull to the correction unit 242.

[0142] On the basis of foreground depth images, which are the luminance components of the input depth-related images of the plurality of virtual cameras, the correction unit 242 corrects the convex hull supplied from the convex hull generation unit 241, to reconstruct the 3D model of the foreground, like the correction unit 102 in FIG. 4. The correction unit 242 supplies the 3D model of the foreground to the mesh generation unit 103.

[0143] The reconstruction unit 33 according to the third embodiment generates a 3D model of the foreground without the use of any foreground information as described above.

Example of a Convex Hull

[0144] FIG. 13 is a diagram showing an example of a convex hull to be generated by the convex hull generation unit 241 shown in FIG. 12.

[0145] In FIG. 13, the same components as those shown in FIG. 5 are denoted by the same reference numerals as those used in FIG. 5. The explanations that have already been made will not be repeated.

[0146] In the example shown in FIG. 13, the viewpoints A through D are arranged so as to surround the foreground object 121, as in the example shown in FIGS. 5 through 9. In this case, the convex hull generation unit 241 generates the 3D region included in all the angles of view 141 through 144 of the viewpoints A through D as a convex hull 260.

[0147] Specifically, the convex hull generation unit 241 first generates a convex hull that is a 3D region included in the angle of view 141 of the viewpoint A. The convex hull generation unit 241 then generates a new convex hull that is a 3D region included in the angle of view 142 of the viewpoint B in the generated convex hull. After that, the convex hull is updated by sequentially using the angle of view 143 of the viewpoint C and the angle of view 144 of the viewpoint D in a manner similar to the above, so that the convex hull 260 is finally generated. This convex hull 260 includes the object 121.

[0148] Like the correction unit 102 in FIG. 4, the correction unit 242 corrects the convex hull 260 on the basis of the foreground depth images, to generate a convex hull of the same shape as the object 121 as a 3D model of the foreground.

[0149] Note that a decoding process according to the third embodiment is similar to the decoding process shown in FIG. 10, except that the process in step S33 is replaced with a process to be performed by the convex hull generation unit 241 to generate a convex hull, and the process in step S34 is replaced with a process to be performed by the correction unit 242 to generate a 3D model of the foreground by correcting the convex hull.

FOURTH EMBODIMENT

Example Configuration of the Reconstruction Unit

[0150] A fourth embodiment of the image processing system to which the present disclosure is applied differs from the first embodiment in the method of reconstructing a 3D model. Specifically, the configuration of the fourth embodiment of an image processing system to which the present disclosure is applied is similar to the configuration shown in FIG. 1, except for the configuration of the reconstruction unit 33. Therefore, explanation of the components other than the reconstruction unit 33 will not be unnecessarily repeated below.

发表评论

电子邮件地址不会被公开。 必填项已用*标注