JP2018113054A - Image processing device, image processing method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing device capable of easily generating an image of clothing used for virtual try-on, an image processing method, and a program.SOLUTION: An image processing device displays: a first three-dimensional model regarding a subject and a first control point on the first three-dimensional model; a second three-dimensional model that is the first three-dimensional model subjected to deformation and a second control point on the second three-dimensional model; a clothing region of the subject wearing clothing and an apex in the clothing region; and a clothing image generated from the first control point, the second control point corresponding to the first control point, and an apex in the clothing region corresponding to the first control point.SELECTED DRAWING: Figure 3

Description

  Embodiments described herein relate generally to an image processing apparatus, an image processing method, and a program.

  In recent years, a technique capable of acquiring, for example, a three-dimensional shape and a texture map (hereinafter referred to as a 3DCG model) of the whole body by scanning with a range finder has been developed.

  According to this technique, a 3DCG model of the clothing portion can be generated by scanning a subject wearing the clothing (for example, a mannequin). When the 3DCG model of the clothing part generated in this way is displayed on a screen including a user, for example, the user virtually tries on the clothing (hereinafter referred to as virtual) without actually trying it on. It can be expressed as “try-on”.

JP 2011-180790 A

  By the way, the state (orientation, posture, body shape, etc.) of the user who performs the virtual fitting as described above is various. For this reason, it is preferable to generate a 3DCG model corresponding to each of these user states. However, it is difficult to prepare all such 3DCG models from the viewpoint of cost (expense) and time required to generate the 3DCG model.

  Therefore, the problem to be solved by the present invention is to provide an image processing apparatus, an image processing method, and a program capable of easily generating an image of clothes used for virtual fitting.

  The image processing apparatus according to the embodiment displays a first three-dimensional model related to a subject and a first control point on the first three-dimensional model, and a second modified version of the first three-dimensional model. Displaying a three-dimensional model and a second control point on the second three-dimensional model; displaying a clothing region of the subject wearing the clothing; and a vertex in the clothing region; and the first control point A clothing image generated from the second control point corresponding to the first control point and the vertex in the clothing region corresponding to the first control point is displayed.

FIG. 1 is a diagram for explaining a configuration of an image processing system according to the first embodiment. FIG. 2 is a diagram for explaining the configuration of the image processing system. FIG. 3 is a block diagram mainly illustrating an example of a functional configuration of the image processing apparatus. FIG. 4 is a flowchart illustrating an example of a processing procedure of the image processing apparatus. FIG. 5 is a diagram illustrating an example of the 3D model before deformation. FIG. 6 is a diagram illustrating an example of a clothing region. FIG. 7 is a diagram illustrating an example of a clothing image acquired from a clothing captured image. FIG. 8 is a diagram illustrating an example of control points set on the pre-deformation 3D model. FIG. 9 is a diagram illustrating an example of mesh data. FIG. 10 is a diagram for explaining processing when calculating the influence degree. FIG. 11 is a diagram for explaining the distance between the target control point and the target vertex calculated as the degree of influence. FIG. 12 is a diagram illustrating an example of the post-deformation 3D model. FIG. 13 is a diagram illustrating an example of control points set on the post-deformation 3D model. FIG. 14 is a diagram illustrating an example of post-deformation mesh data. FIG. 15 is a diagram illustrating an example of a clothing image generated by the clothing image generation unit. FIG. 16 is a diagram illustrating an example of the post-deformation 3D model in which the position of the arm is changed. FIG. 17 is a block diagram mainly illustrating an example of a functional configuration of the image processing apparatus according to the second embodiment. FIG. 18 is a flowchart illustrating an example of a processing procedure of the image processing apparatus. FIG. 19 is a diagram illustrating an example of a clothing captured image. FIG. 20 is a diagram illustrating an example of a clothing area. FIG. 21 is a diagram for explaining boundary information. FIG. 22 is a diagram for explaining the distance between the vertex calculated as the degree of influence and the control point. FIG. 23 is a block diagram mainly illustrating an example of a functional configuration of the image processing apparatus according to the third embodiment. FIG. 24 is a flowchart illustrating an example of a processing procedure of the image processing apparatus. FIG. 25 is a block diagram mainly showing a functional configuration of the image processing apparatus according to the fourth embodiment. FIG. 26 is a flowchart illustrating an example of a processing procedure of the image processing apparatus. FIG. 27 is a diagram for describing a clothing image generated in the present embodiment. FIG. 28 is a diagram for explaining the opacity of the first image and the second image in the boundary region.

  Hereinafter, each embodiment will be described with reference to the drawings.

(First embodiment)
First, an example of the configuration of an image processing system including an image processing apparatus according to the first embodiment will be described with reference to FIGS. 1 and 2. As shown in FIGS. 1 and 2, the image processing system includes a first imaging unit 100, a second imaging unit 200, an image processing device 300, and the like.

  As illustrated in FIG. 1, the first imaging unit 100 captures an image of a subject 402 placed on a turntable 401 that can rotate at an arbitrary angle, for example, and includes an image including the subject 402 (hereinafter, the depth of the subject 402). This is an imaging device for outputting (denoted as an image). The depth image is also called a distance image, and is an image in which the distance from the first imaging unit 100 is defined for each pixel. For example, a depth sensor is used as the first imaging unit 100.

  As shown in FIG. 2, the second imaging unit 200 is placed on the turntable 401 described above, images the subject 402 wearing the clothing 403, and includes an image including the subject 402 wearing the clothing 403 (hereinafter, This is an imaging device for outputting clothing captured images. The clothing captured image is a bitmap image, and is an image in which pixel values indicating the color, brightness, and the like of the subject 402 (clothes 403 worn by the subject) are defined for each pixel. As the second imaging unit 200, for example, a general-purpose camera or the like is used.

  Here, in the present embodiment, the subject 402 imaged by the first imaging unit 100 and the second imaging unit 200 is an object to try on clothes. 1 and 2, it is assumed that the subject 402 is a mannequin that imitates the shape of a human body, for example. In the following description, the subject 402 is described as a mannequin. However, the subject 402 may be a mannequin that imitates the shape of a pet, such as a dog or a cat, and other objects, such as a person, It may be a living organism such as a pet.

  The depth image of the subject 402 output by the first imaging unit 100 and the clothing captured image output by the second imaging unit 200 are acquired by the image processing device 300. The image processing apparatus 300 uses the acquired depth image and captured clothing image of the subject 402 to change the subject 402 in a state (for example, orientation, posture, or body shape) different from the subject 402 wearing the clothing 403 described above. Has a function of generating an image (hereinafter referred to as a clothing image) representing the state of the clothing 403 when the wearer is worn. The clothing image generated by the image processing apparatus 300 in this way is accumulated (stored) in the image processing apparatus 300.

  The clothing image stored in the image processing apparatus 300 according to the present embodiment is used for virtual try-on. Specifically, for example, a clothing image stored in the image processing apparatus 300 on an image including the user imaged by the imaging unit on a display device provided at a position facing a user in a store selling the clothing 403. By displaying a composite image in which (the image of clothes 403) is superimposed, the user can virtually try on the clothes 403 (hereinafter referred to as virtual try-on).

  Although omitted in FIGS. 1 and 2, the image processing system may further include a server device that is communicably connected to the image processing device 300 via a network, for example. In this case, the clothing image generated by the image processing device 300 may be stored (stored) in the server device.

  FIG. 3 is a block diagram mainly showing a functional configuration of the image processing apparatus 300 according to the present embodiment. As shown in FIG. 3, the image processing apparatus 300 includes a 3D model acquisition unit 301, a 3D model projection unit 302, a control point setting unit 303, a control point change unit 304, a captured captured image acquisition unit 305, a clothing region extraction unit 306, A mesh data generation unit 307, an influence calculation unit 308, a mesh data transformation unit 309, a clothing image generation unit 310, a presentation processing unit 311 and a clothing image storage unit 312 are included.

  In the present embodiment, some or all of these units 301 to 311 are realized by causing a computer such as a CPU provided in the image processing apparatus 300 to execute a program, that is, by software. Part or all of the units 301 to 311 may be realized by hardware such as an IC (Integrated Circuit), or may be realized as a combined configuration of software and hardware. In the present embodiment, it is assumed that the clothing image storage unit 312 is stored in a storage device such as an HDD (Hard Disk Drive) provided in the image processing apparatus 300. Note that the clothing image storage unit 312 may be provided in, for example, an external device (server device) connected to the image processing apparatus 300 so as to be communicable.

  The 3D model acquisition unit 301 acquires a depth image of the subject 402 output from the first imaging unit 100 described above. The 3D model acquisition unit 301 acquires (generates) a three-dimensional model (hereinafter referred to as a 3D model) of the subject 402 by modeling the subject 402 based on the acquired depth image of the subject 402. Here, the 3D model of the subject 402 acquired by the 3D model acquisition unit 301 is a model representing the state (orientation, posture, body shape, etc.) of the subject 402 imaged by the first imaging unit 100.

  Further, the 3D model acquisition unit 301 acquires a 3D model different from the 3D model by deforming the acquired 3D model, for example.

  In the following description, the 3D model (first three-dimensional model) acquired based on the depth image of the subject 402 is converted into a 3D model before deformation, and the 3D model (first image acquired by deforming the 3D model before deformation). 2 three-dimensional model) is referred to as a post-deformation 3D model. The post-deformation 3D model includes, for example, a 3D model in which the orientation, posture, or body shape of the pre-deformation 3D model is changed.

  As described above, the 3D model (pre-deformation 3D model and post-deformation 3D model) acquired by the 3D model acquisition unit 301 includes a plurality of vertices, and each of the plurality of vertices has a three-dimensional position of the vertex. (Coordinates) is added. It is assumed that a number (hereinafter referred to as a vertex number) for identifying the vertex is assigned to each of the plurality of vertices constituting the 3D model. Note that the same vertex number is assigned to the corresponding vertex in the pre-deformation 3D model and the post-deformation 3D model (that is, a vertex at the same structural position in the 3D model).

  The 3D model projection unit 302 projects the 3D model (pre-deformation 3D model and post-deformation 3D model) acquired by the 3D model acquisition unit 301 onto, for example, a screen surface. In this case, the 3D model projection unit 302 projects the 3D model in accordance with the viewpoint position and the angle of view of the second imaging unit 200 described above (that is, the viewpoint position and the angle of view capturing the clothing captured image).

  The control point setting unit 303 sets a plurality of control points (first control points) on the pre-deformation 3D model acquired by the 3D model acquisition unit 301. In this case, the control point setting unit 303 refers to the pre-deformation 3D model projected on the screen surface by the 3D model projection unit 302 and sets control points from among a plurality of vertices constituting the pre-deformation 3D model ( select.

  Similarly, the control point setting unit 303 sets a plurality of control points (second control points) on the transformed 3D model acquired by the 3D model acquisition unit 301. In this case, the control point setting unit 303 sets (selects) a control point corresponding to the control point set on the pre-deformation 3D model among the plurality of vertices constituting the post-deformation 3D model.

  The control point changing unit 304 changes the control point set by the control point setting unit 303, for example, according to a user operation on the image processing apparatus 300.

  The clothing captured image acquisition unit 305 acquires the clothing captured image output from the second imaging unit 200 described above.

  The clothing region extraction unit 306 extracts a region of the clothing 403 worn by the subject 402 included in the clothing captured image (hereinafter referred to as a clothing region) from the clothing captured image acquired by the clothing captured image acquisition unit 305.

  The mesh data generation unit 307 generates mesh data representing the meshed clothing region by dividing the clothing region extracted by the clothing region extraction unit 306 into a mesh shape. Note that the mesh data generated by the mesh data generation unit 307 has a plurality of vertices represented by meshing the clothing region.

  For example, when the pre-deformation 3D model is deformed (that is, when the orientation, posture, body shape, or the like of the subject 402 is changed), the influence degree calculation unit 308 is configured to control each of a plurality of control points set on the pre-deformation 3D model. The degree of influence indicating the degree of influence of movement on each of the plurality of vertices of the mesh data generated by the mesh data generation unit 307 (that is, the degree of influence of the control point on the vertex) calculate. The influence calculation unit 308 includes a plurality of control points set on the 3D model before deformation by the control point setting unit 303 and a plurality of vertices included in the mesh data generated by the mesh data generation unit 307. The degree of influence is calculated for each combination.

  The mesh data deformation unit 309 includes a plurality of control points set on the 3D model before deformation by the control point setting unit 303, a plurality of control points set on the 3D model after deformation by the control point setting unit 303, and mesh data Based on the mesh data generated by the generation unit 307 and the influence degree calculated by the influence degree calculation unit 308, a process of deforming the mesh data is executed. The details of the processing executed by the mesh data deforming unit 309 will be described later, but the mesh data deforming unit 309 controls the control set on the post-deformation 3D model from a plurality of control points set on the pre-deformation 3D model. Based on the movement amount to the point, the mesh data generated by the mesh data generation unit 307 is deformed. In the following description, the mesh data deformed by the mesh data deforming unit 309 is referred to as post-deformed mesh data.

  The clothing image generation unit 310 generates an image of the clothing 403 (that is, a clothing image) using the clothing captured image (the clothing region extracted from the clothing image) and the deformed mesh data. The clothing image generated by the clothing image generation unit 310 is an image representing the state of the clothing 403 when the deformed 3D model (the subject represented by the orientation, posture, and body shape) wears the clothing 403.

  The clothing image generated by the clothing image generation unit 310 is presented (displayed) to the administrator or the like of the image processing apparatus 300 by the presentation processing unit 311. In addition, the clothing image generated by the clothing image generation unit 310 is stored (accumulated) in the clothing image storage unit 312.

  Next, a processing procedure of the image processing apparatus 300 according to the present embodiment will be described with reference to the flowchart of FIG.

  First, the 3D model acquisition unit 301 acquires a 3D model (pre-deformation 3D model) of the subject 402 based on the depth image of the subject 402 output from the first imaging unit 100 (step S1). Note that although the pre-deformation 3D model is acquired based on the depth image of the subject 402 here, the pre-deformation 3D model is acquired using, for example, a 3D scanner if the structures match. It doesn't matter.

  Here, FIG. 5 shows the pre-deformation 3D model 501 acquired by the 3D model acquisition unit 301. In the example illustrated in FIG. 5, the pre-deformation 3D model 501 is an example of a 3D model acquired based on a depth image of the subject 402 facing substantially in front of the first imaging unit 100, for example. The pre-deformation 3D model 501 is composed of a plurality of vertices. In addition, a three-dimensional position (coordinate) of the vertex is added to each of the plurality of vertices. Furthermore, a vertex number is assigned to each of the plurality of vertices.

  Returning to FIG. 4 again, the clothing captured image acquisition unit 305 acquires the clothing captured image output from the second imaging unit 200 (step S2). This clothing captured image is an image including the subject 402 wearing the clothing 403 in a state (posture and the like) equivalent to that at the time of imaging by the first imaging unit 100 described above.

  Next, the clothing region extraction unit 306 performs, for example, a mask process on the clothing captured image acquired by the clothing captured image acquisition unit 305, thereby changing the clothing captured image from the clothing captured image (that is, the clothing captured image into the clothing captured image). The type of clothes 403 worn by the included subject 402 is extracted (step S3).

  Here, FIG. 6 shows the clothing region 601 extracted by the clothing region extraction unit 306 (that is, the extraction result by the clothing region extraction unit 306). Such extraction of the clothing region 601 may be realized using, for example, a general-purpose image editing tool.

  In addition, by using the clothing region 601 extracted by the clothing region extraction unit 306, a clothing image (that is, the second imaging unit) as illustrated in FIG. 7 can be obtained from the clothing captured image acquired by the clothing captured image acquisition unit 305. The image 701 (image of the clothing 403 in a state worn by the subject 402 taken by the user 200) 701 can be acquired (cut out). The clothing image 701 acquired in this way is stored in the clothing image storage unit 312 for use in the virtual try-on described above.

  Returning to FIG. 4 again, the control point setting unit 303 sets a plurality of control points on the pre-deformation 3D model 501 acquired by the 3D model acquisition unit 301 (step S4). In this case, the 3D model projection unit 302 projects, for example, the pre-deformation 3D model 501 acquired by the 3D model acquisition unit 301 onto the screen surface according to the viewpoint position and the angle of view of the second imaging unit 200. It is assumed that the positional relationship between the first imaging unit 100 and the second imaging unit 200 is calibrated in advance. Thereby, the control point setting unit 303 is a vertex that appears on the forefront when the pre-deformation 3D model 501 is projected on the screen surface among the plurality of vertices constituting the pre-deformation 3D model 501 and is projected. In this case, the vertex located on the outermost periphery of the 3D model before deformation is set (selected) as a control point. In this case, the control point setting unit 303 holds the vertex number assigned to each of the plurality of control points set on the pre-deformation 3D model 501 in the control point setting unit 303.

  Here, it has been described that the control point is selected from all the vertices that make up the 3D model before deformation, but vertices that are susceptible to changes in body orientation, posture, or body shape (ie, physical feature points) It is also possible to select a control point from among the vertices. Specifically, representative vertices (for example, vertices located on the shoulder, elbow, waist, etc.) in the 3D model such as a plurality of orientations, postures, and body shapes may be selected in advance.

  Further, based on the shape of the clothing area extracted by the clothing area extraction unit 306 (that is, the type of the clothing 403), a vertex associated with the clothing 403 is selected in advance, and a control point is selected from the vertex. May be selected. Specifically, if the garment 403 is to be worn on the upper body such as a top, for example, a vertex located on the upper body of the 3D model before deformation may be selected in advance as the vertex associated with the clothing 403. . On the other hand, if the garment 403 is to be worn on the lower body such as the bottom, for example, a vertex located on the lower body of the 3D model before deformation may be selected in advance as the vertex associated with the clothing 403.

  Here, FIG. 8 shows an example of control points set on the pre-deformation 3D model 501 shown in FIG. In FIG. 8, only one control point is given a reference for convenience, but a plurality of control points including the control point 501a are set (arranged) on the 3D model 501 before deformation. In the example illustrated in FIG. 8, the garment 403 is worn on the upper body, and an example is shown in which the control points are set from the vertices located on the upper body of the 3D model before deformation.

  Note that the pre-deformation 3D model 501 in which a plurality of control points are set by the control point setting unit 303 as shown in FIG. 8 is presented to, for example, an administrator. Thereby, the administrator can confirm a plurality of control points set on the 3D model 501 before deformation. Here, for example, when the administrator confirms a plurality of control points and, as a result, desires to change the control points, the control point changing unit 304 changes the control points (for example, control points according to the operation of the administrator). Move, add and delete). According to this, it is possible to set the vertex intended by the user on the 3D model 501 before deformation.

  Returning to FIG. 4 again, the mesh data generation unit 307 converts the clothing region 601 extracted by the clothing region extraction unit 306 into a mesh (that is, divides it into a mesh shape), thereby generating mesh data ( That is, mesh data representing meshed clothing region 601 is generated (step S5).

  Here, FIG. 9 shows the mesh data 801 generated by the mesh data generation unit 307. In FIG. 9, only one vertex is given a reference for convenience, but the mesh data 801 has a plurality of vertices including a vertex 801a. The plurality of vertices are vertices of small areas in the clothes area 601 divided in a mesh shape. Such generation of mesh data may be realized using, for example, a general-purpose CG (Computer Graphics) tool.

  Returning to FIG. 4 again, the influence calculation unit 308 includes the mesh data 801 generated by the mesh data generation unit 307 and each of the plurality of control points set on the pre-deformation 3D model 501 by the control point setting unit 303. The degree of influence for each combination of a plurality of vertices (that is, the degree of influence of each control point on each vertex) is calculated (step S6).

  Here, when calculating the degree of influence of one control point (hereinafter referred to as a target control point) 501a among a plurality of control points on one vertex (hereinafter referred to as a target vertex) 801a among a plurality of vertices. The process will be specifically described.

  In this case, as shown in FIG. 10, the influence calculation unit 308 is generated by the mesh data generation unit 307 with respect to the image 901 obtained by projecting the 3D model 501 before deformation onto the screen surface by the 3D model projection unit 302 described above. The image 902 is generated by superimposing the mesh data (that is, mesh data representing the meshed clothing region 601). Note that the image 902 is an image representing a state in which the subject 402 wears the clothes 403.

  Next, the influence calculation unit 308 uses the area of the pre-deformation 3D model 501 in the image 902 (that is, the area obtained by projecting the pre-deformation 3D model 501 on the screen surface) as the influence of the target control point 501a on the target vertex 801a. The distance between the target control point 501a and the target vertex 801a in the sum area 903 with the area of the mesh data 801 (that is, the clothes area 601 represented by the mesh data 801) is calculated.

  Here, with reference to FIG. 11, the distance between the target control point 501a and the target vertex 801a calculated as the impact level by the impact level calculation unit 308 will be described in detail. FIG. 11 is an enlarged view of the region 903 (the sum region of the region of the 3D model 501 before deformation and the region of the mesh data 801) shown in FIG.

  In the example illustrated in FIG. 11, it is assumed that the target control point 501a is set at a position inside the arm of the 3D model 501 (region) before deformation. On the other hand, it is assumed that the target vertex 801a is arranged near the center of the mesh data 801.

  In this case, the influence calculation unit 308 calculates the shortest path 1001 from the target vertex 801a passing through the area 903 to the target control point 501a, and based on the calculated shortest path, the target control point 501a and the target vertex 801a. The distance between is calculated.

  When the distance between the target control point 501a and the target vertex 801a is calculated as an influence degree, the influence degree of the target control point 501a on the target vertex 801a increases as the value (that is, the distance) decreases. Represents that.

  Here, when the path from the target vertex 801a to the target control point 501a bends as in the example shown in FIG. 11, the value based on the bend angle is multiplied or added, based on the path. The calculated distance may be increased (that is, the influence of the target control point 501a on the target vertex 801a is reduced).

  Further, when the path from the target vertex 801a to the target control point 501a intersects the skeleton (bone) of the 3D model (subject 402) before deformation, the distance calculated based on the path may be increased. Good. It is assumed that the skeleton (position) of the 3D model before deformation is estimated from the position of the joint in the 3D model before deformation.

  In the present embodiment, by calculating the distance between the target control point 501a and the target vertex 801a as an influence level, for example, the control point set at the arm position of the 3D model 501 and the torso position are set. It is possible to prevent the set control points from affecting each other.

  Returning to FIG. 4 again, the 3D model acquisition unit 301 acquires the post-deformation 3D model by deforming the pre-deformation 3D model 501 acquired in step S1 (step S7). Note that the post-deformation 3D model is generated by, for example, deforming (changing) at least one of the orientation, posture, and body shape of the pre-deformation 3D model.

  Here, FIG. 12 shows the post-deformation 3D model 502 acquired by the 3D model acquisition unit 301. In the example shown in FIG. 12, the post-deformation 3D model 502 is an example of a 3D model in which the orientation of the pre-deformation 3D model 501 shown in FIG. 5 is changed by about 10 degrees. The post-deformation 3D model 502 includes a plurality of vertices similar to the pre-deformation 3D model 501. In the pre-deformation 3D model and the post-deformation 3D model, the same vertex numbers are assigned to the corresponding vertices (that is, vertices at the same structural position).

  In addition, if correspondence of the vertex (vertex number assigned to) with the 3D model before deformation can be taken, for example, the subject 402 whose orientation has been changed by rotating the turntable 401 described above is the first subject. A post-deformation 3D model may be acquired (generated) based on a depth image output by imaging by the imaging unit 100.

  Returning to FIG. 4 again, the control point setting unit 303 sets a plurality of control points on the post-deformation 3D model 502 acquired by the 3D model acquisition unit 301 (step S8). In this case, the control point setting unit 303 sets a control point corresponding to the control point set on the pre-deformation 3D model 501 among a plurality of vertices constituting the post-deformation 3D model 502 (that is, the pre-deformation 3D). The vertex corresponding to the vertex set as the control point on the model 501 is set as the control point). Specifically, the control point setting unit 303 is assigned to each of the vertex numbers held in the control point setting unit 303 in step S4 described above (that is, each of the plurality of control points set on the 3D model 501 before deformation). The vertex having the same vertex number as the control point is set as the control point.

  Here, FIG. 13 shows an example of control points set on the post-deformation 3D model 502 shown in FIG. In FIG. 13, for convenience, only one control point is provided with a reference symbol, but a plurality of control points including the control point 502a are set (arranged) on the transformed 3D model 502. Note that the control point 502a set at the same position (left arm) as the control point 501a set on the pre-deformation 3D model 501 shown in FIG. 8 is the control point corresponding to the control point 501a (that is, The control point assigned the same vertex number as the control point 501a).

  Returning to FIG. 4 again, the mesh data deforming unit 309 includes a plurality of control points set on the 3D model before deformation in step S4, and the degree of influence calculated in step S7 (that is, the influence of each control point on each vertex). Degree) and a plurality of control points set on the 3D model after deformation in step S8, the mesh data generated in step S5 is deformed (step S9). In this case, the mesh data deforming unit 309 performs, for example, weighted averaging of movement amounts (vectors) of corresponding control points set on the pre-deformation 3D model and the post-deformation 3D model by the degree of influence, so that a plurality of mesh data has Determine (change) the position of each vertex. By executing such processing, as shown in FIG. 14, mesh data (deformed mesh data) 802 obtained by modifying the mesh data 801 shown in FIG. 9 is generated.

  For example, a Rigid deformation technique can be used as the mesh data deformation algorithm. As the distance used in this method, the above-described influence degree may be used.

  Hereinafter, the deformation process of mesh data when the Rigid Deformation method is used will be described in detail. Here, a case where the position of a certain vertex j among a plurality of vertices included in the mesh data is determined will be described. Note that j is an index of a vertex included in the mesh data.

First, the mesh data deformation unit 309 calculates w _i for all control points i based on the following equation (1). Note that i is an index of a control point set on the 3D model 501 before deformation.

In this equation (1), D _i is the distance from the vertex j to the control point i (that is, the degree of influence of the control point i on the vertex j). Moreover, (alpha) in Formula (1) is a constant, for example, can be 1 or 2 grade | etc.,.

  Next, the mesh data deforming unit 309 calculates μ based on the following equation (4).

Next, the mesh data deformation unit 309 calculates A _i based on the following equation (6).

  In the present embodiment, the two-dimensional coordinates in the deformed mesh data are calculated for all of the plurality of vertices of the mesh data using the above-described formulas (1) to (10). The mesh data 801 shown in FIG. 9 can be transformed into post-deformation mesh data 802 by setting the positions of the vertices of the mesh data to the two-dimensional coordinates calculated in this way.

  It is assumed that the connection relationship between the vertices for configuring the mesh surface in the mesh data does not change in the process of step S9. Further, as described above, all the vectors used in the mesh data transformation algorithm are assumed to be row vectors.

  As described above, when the path from the target vertex 801a to the target control point 501a intersects with the skeleton of the 3D model before deformation, when the distance calculated based on the path is increased, the area close to the skeleton In some cases, distortion such as turning over the mesh in the mesh data 802 after deformation may occur. In this case, by fixing the vertices near the contour of the post-deformation mesh data 802 and applying a Laplacian filter to each vertex (coordinates) of the post-deformation mesh data 802, the post-deformation mesh data 802 is distorted. May be removed.

  Next, the clothing image generation unit 310 generates an image (clothing image) of the clothing 403 using the clothing photographed image and the deformed mesh data 802 (step S10).

  Here, the shape of the clothing region 601 represented by the mesh data (hereinafter referred to as pre-deformation mesh data) 801 generated in step S5 described above matches the shape of the clothing 403 included in the clothing captured image. For this reason, colors corresponding to a plurality of vertices included in the pre-deformation mesh data 801 can be identified from the clothes photographed image.

  For this reason, the clothing image generation unit 310 generates a clothing image 702 as shown in FIG. 15, for example, by adding a corresponding color to each of a plurality of vertices included in the post-deformation mesh data 802. Can do. Note that the clothing image 702 shown in FIG. 15 is obtained by deforming the clothing image 701 shown in FIG. 7 described above, and the state of the clothing 403 when the 3D model 502 after the deformation wears the clothing 403 (that is, the second image 702). The image of the subject 402 imaged by the imaging unit 200 of FIG.

  Next, the presentation processing unit 311 presents the clothing image 702 generated by the clothing image generation unit 310 (step S11). As a result, the administrator can confirm the clothing image 702 by the clothing image generation unit 310.

  Here, as a result of confirming the clothing image 702, if the administrator determines that the clothing image 702 is not appropriate (that is, the clothing image 702 needs to be corrected), the above-described 3D model 501 before deformation described above. It is possible to instruct change of the control point set above (for example, movement, addition and deletion of the control point).

  In this case, it is determined whether or not an instruction to change the control point set on the pre-deformation 3D model 501 has been issued (step S12).

  If it is determined that the change of the control point has been instructed (YES in step S12), for example, the control point is changed according to the operation of the administrator, and then the process returns to step S6 described above and the process is repeated.

  On the other hand, when it is determined that the change of the control point has not been instructed (NO in step S12), that is, when it is not necessary to correct the clothing image 702, the clothing image 702 is stored in the clothing image storage unit 312. (Step S13). In this case, the clothing image 702 is stored in association with the state (orientation, posture, body shape, etc.) of the post-deformation 3D model 502, for example.

  Here, for example, by acquiring the post-deformation 3D model 502 in which the orientation of the pre-deformation 3D model 501 is changed, a clothing image 702 representing a state in which the orientation of the subject 402 is changed from that of the clothing 403 worn by the subject 402 is generated. As described above, for example, by acquiring a deformed 3D model 503 in which the position of the arm is changed as illustrated in FIG. 16, a clothes image in which the position of the sleeve is changed from the clothes 403 worn by the subject 402 is generated. It is also possible. As described above, in the present embodiment, by acquiring (generating) various post-deformation 3D models in which the orientation, posture, and body shape of the pre-deformation 3D model 501 are changed, subjects in various states (orientations, postures, or body shapes) are obtained. When the user wears the clothes 403, it is possible to generate a clothes image representing the state of the clothes 403. Specifically, clothing images with orientations of ± 5 degrees and ± 10 degrees from the angle (for example, rotation angle 0 degree) of the 3D model 501 before deformation, clothing images with different ways of opening arms, and fitting to different body shapes It is possible to generate a clothing image or the like in the finished state.

  As described above, in the present embodiment, mesh data is obtained by acquiring the pre-deformation 3D model (first three-dimensional model) 501 of the subject 402 and dividing the clothing region 601 extracted from the captured clothing image into a mesh shape. A post-deformation 3D model (second three-dimensional model) 502 is obtained by generating 801 and deforming the pre-deformation 3D model 501. Furthermore, in the present embodiment, a plurality of control points (second control points) set on the post-deformation 3D model 502 from a plurality of control points (first control points) set on the pre-deformation 3D model 501. The mesh data 801 is deformed based on the movement amount to (), and a clothes image is generated using the clothes photographed image and the deformed mesh data 802.

  That is, in this embodiment, with such a configuration, the second imaging unit 200 only captures a specific orientation, posture, and body-shaped subject (for example, a mannequin) 402 wearing a garment 403. When a posture and body subject wears the clothing 403, a clothing image representing the state of the clothing 403 (that is, a clothing image obtained by two-dimensionally deforming the image 701 of the clothing 403 worn by the subject 402) is generated. It becomes possible. For this reason, in the present embodiment, it is possible to reduce the cost and time required for generating a clothing image, and it is possible to easily generate a clothing image used for virtual fitting.

  Further, in this embodiment, when the pre-deformation 3D model 501 is deformed, the movement of each of the plurality of control points set on the pre-deformation 3D model 501 is moved to each of the plurality of vertices of the mesh data 801. The degree of influence representing the degree of influence is calculated, and the mesh data is transformed using the degree of influence. Note that the degree of influence calculated in the present embodiment is the area of the untransformed 3D model 501 when the mesh data 801 is superimposed on the untransformed 3D model 501 so that the subject 402 represents the state of wearing the clothes 403. And the distance between each control point and each vertex in the sum area of the clothes area 601 represented by the mesh data 801. In this embodiment, the mesh data can be deformed so that the arm control points and the torso control points set on the 3D model, for example, do not affect each other, and thus more appropriately. Mesh data can be transformed, and as a result, a highly accurate clothing image 702 can be generated.

  Further, in the present embodiment, for example, vertices located on the outer periphery of the pre-deformation 3D model 501 and corresponding to physical feature points in the pre-deformation 3D model 501 are defined as control points on the pre-deformation 3D model 501. By setting to, mesh data can be deformed so as to appropriately reflect the movement of control points between the pre-deformation 3D model 501 and the post-deformation 3D model 502.

  When the image processing apparatus 300 according to the present embodiment accumulates clothing images of various patterns (orientations, postures, and body shapes), when the user performs virtual try-on in the store as described above, for example, imaging The body type parameter of the user captured by the unit can be acquired, and a clothing image associated with a body type parameter similar to the body type parameter can be selected. Thereby, it is possible to display a highly accurate composite image representing a state in which the user wears clothes (a state in which the clothes are fitted to the user's orientation, posture, and body shape).

  In the present embodiment, it has been described that clothes images of various patterns are stored. For example, for each garment 403, a 3D model of the subject 402 wearing the garment 403 (that is, a 3D model before deformation) 501 and only the captured clothing image are stored, and the 3D model after deformation corresponding to the user's state is acquired at the time of virtual fitting as described above, and a clothing image that fits the orientation, posture, and body shape of the user is generated. It doesn't matter. That is, the generation of the clothing image in the present embodiment may be executed at the time of virtual try-on in the store.

(Second Embodiment)
Next, a second embodiment will be described. FIG. 17 is a block diagram mainly showing a functional configuration of the image processing apparatus according to the present embodiment. The same parts as those in FIG. 3 described above are denoted by the same reference numerals, and detailed description thereof is omitted. Here, parts different from FIG. 3 will be mainly described.

  The configuration of the image processing system including the image processing apparatus according to the present embodiment is the same as that of the first embodiment described above, and will be described with reference to FIGS. 1 and 2 as appropriate.

  As illustrated in FIG. 17, the image processing apparatus 1100 according to the present embodiment includes a boundary information acquisition unit 1101 and a depth information calculation unit 1102.

  Here, for example, when the garment 403 worn by the subject 402 is a garment with loose arms or torso (loose), there is a possibility that the arm and the torso are integrated in the garment region extracted from the captured clothing image. There is. When the arm and the torso are integrated in this way, for example, the shortest distance between the control point set at the position of the arm and the apex arranged on the torso is shortened (that is, the Therefore, when the mesh data is deformed, the body may be distorted in accordance with the change in the arm position.

  Therefore, in the present embodiment, the boundary information acquisition unit 1101 acquires boundary information indicating the boundary of the clothing region extracted by the clothing region extraction unit 306. Note that the boundary information acquired by the boundary information acquisition unit 1101 is the distance between each of a plurality of control points set on the 3D model before deformation and each of a plurality of vertices included in the mesh data (described later). That is, it is used when calculating the degree of influence of each control point on each vertex.

  In addition, the mesh data of the clothing area in the first embodiment described above is two-dimensional data. For this reason, when mesh data is deformed, for example, the meshes of the arm and the torso may overlap each other. When a clothing image is generated using such mesh data, the clothing image may not be drawn correctly.

  Therefore, in the present embodiment, the depth information calculation unit 1102 calculates depth information corresponding to each of a plurality of vertices included in the mesh data generated by the mesh data generation unit 307. The depth information (coordinates) corresponding to each of the plurality of vertices is added to the three-dimensional position (coordinates) added to the control points (vertices) set on the 3D model after deformation by the control point setting unit 303. It is calculated based on the included depth information (information indicating the position of the control point in the depth direction with respect to the second imaging unit 200). The depth information calculated by the depth information calculation unit 1102 is used when a clothing image is generated using the deformed mesh data as will be described later.

  Next, a processing procedure of the image processing apparatus 1100 according to the present embodiment will be described with reference to the flowchart of FIG.

  First, steps S21 to S23 corresponding to the steps S1 to S3 shown in FIG. 4 are executed.

  Next, the boundary information acquisition unit 1101 acquires boundary information indicating the boundary (region boundary) of the clothing region extracted from the clothing captured image by the clothing region extraction unit 306 (step SS24).

  Specifically, when a clothing area is extracted in step S23 described above, the clothing area is presented to an administrator or the like. As a result, the administrator can check the clothing region (that is, the extraction result) extracted by the clothing region extraction unit 306.

  Here, FIG. 19 shows an example of a clothing captured image acquired by the clothing captured image acquisition unit 305. FIG. 20 shows an example of a clothing region extracted from the clothing captured image shown in FIG. In the clothing region shown in FIG. 20, the arms and the torso are integrated. In such a clothing region (which is meshed mesh data), the shortest distance between the control point set at the position of the arm and the apex placed on the torso, for example, in the influence calculation process described later Is calculated to be short, and the degree of influence is not appropriate.

  In this case, as shown in FIG. 21, for example, the administrator can designate a boundary (representing boundary information) on the presented clothing area. In the example shown in FIG. 21, boundaries 1201 and 1202 are designated at both sides of the clothing area.

  Thereby, in step S24, the boundary information acquisition part 1101 can acquire the boundary information designated by the administrator as described above.

  The boundary information acquired by the boundary information acquisition unit 1101 in this way is added (given) to the clothing area as attribute information.

  When the process of step S24 is executed, the processes of steps S25 and S26 corresponding to the processes of steps S4 and S5 shown in FIG. 4 are executed.

  Next, the depth information calculation unit 1102 calculates depth information corresponding to each of a plurality of vertices included in the mesh data generated by the mesh data generation unit 307 (step S27).

  Specifically, the depth information calculation unit 1102 has, for example, the mesh data in the case where the mesh data is superimposed on the 3D model before deformation so that the subject 402 represents the state of wearing the clothes 403 as described above. The position of the vertex in the depth direction from the three-dimensional position (coordinates) added to the post-deformation 3D model corresponding to the control point (ie, the apex constituting the pre-deformation 3D model) located in the vicinity of each of the vertices Coordinates (that is, depth information) are calculated (interpolated). Thus, the depth information corresponding to each of the plurality of vertices included in the mesh data calculated by the depth information calculation unit 1102 is added to the vertices.

  Next, the influence calculation unit 308 includes each of a plurality of control points set on the 3D model before deformation by the control point setting unit 303 and each of a plurality of vertices included in the mesh data generated by the mesh data generation unit 307. The degree of influence for each combination (the degree of influence of each control point on each vertex) is calculated (step S28).

  In this case, as described in the first embodiment, the influence degree calculation unit 308 is based on the shortest path from each vertex to each control point, and the influence degree of each control point with respect to each vertex (that is, each The distance between the vertex and each control point) is calculated. In the present embodiment, as shown in FIG. 22, the shortest path in this case is a path that does not cross the boundaries 1201 and 1202 indicated by the boundary information added to the clothes area (mesh data) (for example, a path) 1301).

  When the process of step S28 is executed, the processes of steps S29 to S31 corresponding to the processes of steps S7 to S9 shown in FIG. 4 are executed.

  Next, the clothing image generation unit 310 generates a clothing image by adding a corresponding color to each of a plurality of vertices of the mesh data after deformation (step S32). At this time, the clothing image generation unit 310 generates a clothing image with a correct depth relationship by using, for example, a Z buffer together. Specifically, the clothing image generation unit 310 performs hidden surface removal processing and the like based on depth information added to each of a plurality of vertices included in the post-deformation mesh data.

  When the process of step S32 is executed, the processes of steps S33 to S35 corresponding to the processes of steps S11 to S13 shown in FIG. 4 described above are executed. In addition, when it determines with the change of the control point having been instruct | indicated in step S34, it returns to step S27 mentioned above and a process is repeated.

  As described above, in the present embodiment, the depth information corresponding to each vertex is added to each of a plurality of vertices included in the mesh data, and the mesh data (deformed mesh after deformation) is generated when generating the clothing image. By executing the hidden surface removal process based on the depth information added to the data), it is possible to perform drawing with the correct depth relationship, and thus it is possible to generate a highly accurate clothing image.

  Furthermore, in the present embodiment, the boundary information indicating the boundary of the extracted clothing region is acquired, and the appropriate degree of influence is calculated by calculating the distance between each control point and each vertex using the boundary information. As a result, it is possible to generate accurate clothing images according to changes in the pre-deformation 3D model and the post-deformation 3D model.

(Third embodiment)
Next, a third embodiment will be described. FIG. 23 is a block diagram mainly illustrating a functional configuration of the image processing apparatus according to the present embodiment. The same parts as those in FIG. 3 described above are denoted by the same reference numerals, and detailed description thereof is omitted. Here, parts different from FIG. 3 will be mainly described.

  The configuration of the image processing system including the image processing apparatus according to the present embodiment is the same as that of the first embodiment described above, and will be described with reference to FIGS. 1 and 2 as appropriate.

  As illustrated in FIG. 23, the image processing apparatus 1400 according to the present embodiment includes a clothing captured image acquisition unit 1401 and a clothing image generation unit 1402.

  The clothing captured image acquisition unit 1401 acquires a clothing captured image that includes the subject 402 wearing the clothing 403 and is output from the second imaging unit 200. The clothing captured image acquired by the clothing captured image acquisition unit 1401 includes, for example, a clothing captured image (first clothing captured image) including the subject 402 in the first state and a second state different from the first state. A clothing captured image (second clothing captured image) including the subject 402 in the state is included. Specifically, the first clothing captured image includes, for example, a clothing captured image including a subject 402 facing substantially in the front direction with respect to the second imaging unit 200 (and the first imaging unit 100). The captured clothing image includes a captured clothing image including a subject 402 facing in a direction (angle) rotated about 10 degrees from the substantially front direction.

  In the present embodiment, the pre-deformation 3D model acquired by the 3D model acquisition unit 301 is a model representing the first state (for example, orientation, posture, and body shape) of the subject 402. On the other hand, the post-deformation 3D model acquired by the 3D model acquisition unit 301 is a model representing the second state (for example, orientation, posture, and body shape) of the subject 402.

  The clothing image generation unit 1402 generates an image of the clothing 403 using the second clothing captured image acquired by the clothing captured image acquisition unit 1401 and the mesh data (deformed mesh data) deformed by the mesh data deformation unit 309. To do.

  Next, a processing procedure of the image processing apparatus 1400 according to the present embodiment will be described with reference to the flowchart of FIG.

  First, the 3D model acquisition unit 301, based on the depth image of the subject 402 in the first state output from the first imaging unit 100, the 3D model (pre-deformation 3D model) of the subject 402 in the first state. Is acquired (step S41). The pre-deformation 3D model acquired in step S41 is, for example, a model that represents the subject 402 facing substantially in the front direction.

  Next, the garment captured image acquisition unit 1401 acquires the first garment captured image output from the second imaging unit 200. The first clothing captured image is an image including the subject 402 wearing the clothing 403 in a state equivalent to the image capturing by the first image capturing unit 100 described above (that is, the first state). In other words, the first clothing captured image is an image including the subject 402 wearing, for example, the clothing 403 and facing substantially in the front direction.

  In addition to the first clothing captured image, the clothing captured image acquisition unit 1401 includes the second clothing including the subject 402 wearing the clothing 403 in a state different from the first state (that is, the second state). A captured image is acquired. The second clothing captured image is an image including a subject 402 wearing, for example, clothing 403 and facing a direction (angle) rotated about 10 degrees from the front direction.

  That is, the clothing captured image acquisition unit 1401 includes the first clothing captured image including the subject 402 in the first state wearing the clothing 403 and the second clothing imaging including the subject 402 in the second state wearing the clothing 403. An image is acquired (step S42).

  The above-described first and second clothes captured images may be acquired by, for example, the second imaging unit 200 capturing the subject 402 while manually rotating the rotary table 401 described above, or automatically rotate. It may be acquired by controlling the second imaging unit 200 so that the subject 402 placed on the rotated turn table 401 is automatically imaged according to the rotation angle.

  Thereafter, the processes of steps S43 to S49 corresponding to the processes of steps S3 to S9 shown in FIG. 4 are executed.

  Here, the post-deformation 3D model acquired in step S47 is, for example, a model representing the subject 402 (that is, the subject 402 in the second state) facing in a direction (angle) rotated approximately 10 degrees from the front direction. .

  In this case, the mesh data deformed in step S49 (deformed mesh data) represents a region corresponding to the region of the clothing 403 worn by the subject 402 in the second state included in the above-described second clothing captured image. It becomes data.

  For this reason, the clothing image generation unit 1402 applies the post-deformation mesh data (mask image) to the second clothing captured image (that is, executes a mask process), so that the second clothing captured image is extracted from the second clothing captured image. The area of the clothes 403 can be cut out (step S50). Thereby, the clothing image generation unit 1402 generates a clothing image.

  When the process of step S50 is executed, the processes of steps S51 to S53 corresponding to the processes of steps S11 to S13 shown in FIG. 4 described above are executed. In addition, when it determines with the change of the control point having been instruct | indicated in step S52, after changing the said control point according to operation of an administrator, for example, it returns to step S46 and a process is repeated.

  As described above, in the present embodiment, mesh data (deformed mesh data) deformed in the same manner as in the first embodiment described above includes the second subject 402 in the second state wearing the clothing 403. A clothing image is generated by applying it to a clothing image.

  Here, in the first embodiment described above, since a clothing image is generated by adding a corresponding color to each of a plurality of vertices of the mesh data after deformation, the clothing 403 in the clothing image is generated. The color (and pattern) may be different from the actual clothes 403. On the other hand, in this embodiment, since it becomes possible to cut out the region (part) of the clothing 403 from the second clothing captured image, compared with the clothing image in the first embodiment described above, A highly accurate clothing image can be generated.

  In the present embodiment, as described above, clothing picked-up images of various rotation angles are continuously taken (acquired) by automatically rotating the turntable 401 on which the subject 402 is placed. Thus, clothes images corresponding to the various angles (orientations) can be easily generated. Similarly, a clothing image corresponding to the posture or body shape can be generated by capturing a clothing captured image including the subject 402 having a different posture or body shape.

  In the present embodiment, the area of the clothes 403 in the second clothes captured image may not completely match the area of the clothes 403 represented by the above-described deformed mesh data. For this reason, it is also possible to adopt a configuration in which multi-level (for example, 0 to 255) transparency is given mainly to the boundary portion of the mesh data after deformation (that is, the mask image). According to this, even if there is an error between the area of the clothes 403 in the second clothes captured image and the area of the clothes 403 represented by the above-described deformed mesh data, the influence of the error is affected. It becomes possible to reduce.

(Fourth embodiment)
Next, a fourth embodiment will be described. FIG. 25 is a block diagram mainly illustrating a functional configuration of the image processing apparatus according to the present embodiment. The same parts as those in FIGS. 3 and 23 described above are denoted by the same reference numerals, and detailed description thereof is omitted. Here, portions different from those in FIGS. 3 and 23 will be mainly described.

  The configuration of the image processing system including the image processing apparatus according to the present embodiment is the same as that of the first embodiment described above, and will be described with reference to FIGS. 1 and 2 as appropriate.

  The present embodiment is different from the above-described embodiments in that the first embodiment and the third embodiment described above are combined.

  As shown in FIG. 25, the image processing apparatus 1500 according to this embodiment includes a first image generation unit 1501, a second image generation unit 1502, and a clothing image generation unit 1503.

  The first image generation unit 1501 is a functional unit corresponding to the clothing image generation unit 310 described in the first embodiment, and adds a corresponding color to each of a plurality of vertices included in the deformed mesh data. As a result, an image of the clothing 403 (hereinafter referred to as a first image) is generated. Note that the color corresponding to each of the plurality of vertices included in the mesh data after deformation can be specified from the clothes photographed image (first clothes photographed image).

  The second image generation unit 1502 is a functional unit corresponding to the clothes image generation unit 1402 described in the third embodiment, and applies the deformed mesh data to the second clothes photographed image to apply the clothes. 403 images (hereinafter referred to as second images) are generated.

  The clothing image generation unit 1503 generates a clothing image using the first image generated by the first image generation unit 1501 and the second image generated by the second image generation unit 1502.

  Next, a processing procedure of the image processing apparatus 1500 according to the present embodiment will be described with reference to the flowchart of FIG.

  First, steps S61 to S69 corresponding to the steps S41 to S49 shown in FIG. 24 are executed.

  Next, the first image generation unit 1501 generates a first image using the first clothes photographed image and the deformed mesh data (step S70). Note that the first image generation unit 1501 generates a first image by executing the same process as the process of step S10 shown in FIG. 4 described above.

  The second image generation unit 1502 generates a second image using the second clothes photographed image and the deformed mesh data (step S71). Note that the second image generation unit 1502 generates a second image by executing the same process as the process of step S50 shown in FIG. 24 described above.

  Next, the clothing image generation unit 1503 generates a clothing image by mixing the first image generated by the first image generation unit 1501 and the second image generated by the second image generation unit 1502. (Step S72).

  Here, the clothes image (that is, the first image) generated by the processing described in the first embodiment described above is attached with a color corresponding to the mesh data after deformation (each of a plurality of vertices). Since the image is generated by going, the background area or the like is not included in the image, but the color of the clothing 403 in the image may be different from the actual clothing 403 as described above.

  On the other hand, the clothes image (that is, the second image) generated by the processing described in the third embodiment is an image generated by cutting out the area of the clothes 403 from the second clothes photographed image. The color of the clothes 403 in the image has high accuracy. However, if the area of the clothes 403 in the second clothes photographed image and the area of the clothes 403 represented by the deformed mesh data do not completely match, the area of the clothes 403 in the second clothes photographed image May be cut off or a background area may be included in the clothing image.

  For this reason, in the present embodiment, as shown in FIG. 27, the first image is adopted as the area near the contour (hereinafter referred to as the first area) 1601, and the area other than the vicinity of the contour (hereinafter referred to as the second area). The clothing image is generated so that the second image is adopted in the region (1).

  The first area can be defined as an area having a certain number of pixels inward from the outline of the clothing 403 in the first image, for example.

  When the process of step S72 is executed, the processes of steps S73 to S75 corresponding to the processes of steps S11 to S13 shown in FIG. 4 (steps S51 to S53 shown in FIG. 23) are executed. If it is determined in step S74 that a change of the control point has been instructed, for example, the control point is changed in accordance with the operation of the administrator, and then the process returns to step S66 and the process is repeated.

  As described above, in the present embodiment, the first image generated by the process described in the first embodiment described above and the second image generated by the process described in the third embodiment described above. A clothing image is generated by mixing. In the present embodiment, with such a configuration, it is possible to generate a clothing image with high accuracy regarding color without including, for example, a background region.

  In the present embodiment, since the clothing image is generated by using different images (that is, the first image and the second image) for the first region and the second region, respectively, There may be an unnatural state in which the region portion and the second region portion are separated. For this reason, in this embodiment, as shown in FIG. 28, the opacity (or transparency) of the first image and the second image is changed in the boundary region between the first region and the second region. By doing so, it is possible to make the boundary between the images smooth.

  According to at least one embodiment described above, it is possible to provide an image processing apparatus, an image processing system, an image processing method, and a program capable of easily generating an image of clothes used for virtual fitting. .

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and the equivalents thereof.

  DESCRIPTION OF SYMBOLS 100 ... 1st imaging part, 200 ... 2nd imaging part, 300, 1100, 1400, 1500 ... Image processing apparatus, 301 ... 3D model acquisition part, 302 ... 3D model projection part, 303 ... Control point setting part, 304 ... Control point changing unit, 305,1401 ... Clothed captured image acquiring unit, 306 ... Clothing region extracting unit, 307 ... Mesh data generating unit, 308 ... Influence calculating unit, 309 ... Mesh data deforming unit, 310,1402,1503 ... Clothing image generation unit, 311 ... presentation processing unit, 312 ... clothing image storage unit, 1101 ... boundary information acquisition unit, 1102 ... depth information calculation unit, 1501 ... first image generation unit, 1502 ... second image generation unit.

Claims (10)

Displaying a first three-dimensional model relating to a subject and a first control point on the first three-dimensional model;
Displaying a second three-dimensional model obtained by modifying the first three-dimensional model, and a second control point on the second three-dimensional model;
Display the clothes area of the subject wearing the clothes, and the vertex in the clothes area;
Image processing for displaying a clothing image generated from the first control point, the second control point corresponding to the first control point, and a vertex in the clothing region corresponding to the first control point apparatus. Image the subject,
Displaying a first three-dimensional model acquired from the captured image of the subject and a first control point on the first three-dimensional model;
Displaying a second three-dimensional model obtained by modifying the first three-dimensional model, and a second control point on the second three-dimensional model;
Image the subject wearing clothes,
Displaying the clothing region of the subject extracted from the clothing captured image of the subject wearing the imaged clothing, and the vertex in the clothing region;
Image processing for displaying a clothing image generated from the first control point, the second control point corresponding to the first control point, and a vertex in the clothing region corresponding to the first control point apparatus.   The image processing apparatus according to claim 1, wherein the first control point is a point selected from vertices on the displayed first three-dimensional model.   The image processing apparatus according to claim 1, wherein the second control point is a point selected from vertices on the displayed second three-dimensional model.   The image processing apparatus according to claim 1, wherein the vertexes in the clothing area are vertices represented in a meshed clothing area.   The image processing apparatus according to claim 1, wherein the first control point is located on an outer periphery of the displayed first three-dimensional model. An image processing method executed by an image processing apparatus,
Displaying a first three-dimensional model relating to a subject and a first control point on the first three-dimensional model;
Displaying a second three-dimensional model obtained by modifying the first three-dimensional model, and a second control point on the second three-dimensional model;
Displaying a clothing region of the subject wearing the clothing and a vertex in the clothing region;
Displaying a clothing image generated from the first control point, the second control point corresponding to the first control point, and a vertex in the clothing region corresponding to the first control point; An image processing method including: An image processing method executed by an image processing apparatus,
Imaging a subject;
Displaying a first three-dimensional model acquired from the captured image of the subject and a first control point on the first three-dimensional model;
Displaying a second three-dimensional model obtained by modifying the first three-dimensional model, and a second control point on the second three-dimensional model;
Imaging the subject wearing clothes;
Displaying the clothing region of the subject extracted from the clothing captured image of the subject wearing the imaged clothing, and a vertex in the clothing region;
Displaying a clothing image generated from the first control point, the second control point corresponding to the first control point, and a vertex in the clothing region corresponding to the first control point. An image processing method including: A program executed by a computer of an image processing apparatus,
In the computer,
Displaying a first three-dimensional model relating to a subject and a first control point on the first three-dimensional model;
Displaying a second three-dimensional model obtained by modifying the first three-dimensional model, and a second control point on the second three-dimensional model;
Displaying a clothing region of the subject wearing the clothing and a vertex in the clothing region;
Displaying a clothing image generated from the first control point, the second control point corresponding to the first control point, and a vertex in the clothing region corresponding to the first control point; A program for running A program executed by a computer of an image processing apparatus,
In the computer,
Imaging a subject;
Displaying a first three-dimensional model acquired from the captured image of the subject and a first control point on the first three-dimensional model;
Displaying a second three-dimensional model obtained by modifying the first three-dimensional model, and a second control point on the second three-dimensional model;
Imaging the subject wearing clothes;
Displaying the clothing region of the subject extracted from the clothing captured image of the subject wearing the imaged clothing, and a vertex in the clothing region;
Displaying a clothing image generated from the first control point, the second control point corresponding to the first control point, and a vertex in the clothing region corresponding to the first control point. A program to execute and.

Images