JP6467817B2 - Image processing apparatus, image processing method, and program - Google Patents

Description

  The present invention relates to an image processing apparatus, an image processing method, and a program for extracting a contour of a specific area such as a face from an image and performing a process for cutting out the area.

  With the development of image processing technology, the development of high-performance mobile terminals such as so-called smartphone terminals and tablet terminals, or the development of new devices such as digital signage devices, for example, as illustrated in FIG. A technology has been developed that cuts out a face image as exemplified in FIG. 7B from the captured image and provides the face image to various applications such as an application for creating a stamp or an application for displaying an electronic advertisement. Has been.

  In order to cut out a face image from the provided image, it is necessary to extract a face outline from the image. When acquiring an outline for cutting out a face from an image, there is a method of searching for an optimum outline by gradually changing the outline from the initial outline. However, this method has a problem that it takes a long time to search if the number of contour points is large, and it is impossible to accurately acquire a contour having a complicated shape if the number of contour points is small.

To solve this problem, after obtaining the rough contour, Graph Cuts method (hereinafter referred to as "graph cut") method for obtaining a more accurate contour using a technique referred to as is known (e.g. non (Techniques described in Patent Document 1 and Patent Document 1).

  Furthermore, a conventional technique for accurately performing a graph cut on a face area is known (for example, a technique described in Patent Document 2). In this prior art, an initial face area and an initial background area are set in an input image using a plurality of human area patterns that are already stored, and initial values for graph cutting are respectively assigned to these areas. Therefore, it is a technique for editing an image using a graph cut with higher accuracy.

JP 2011-35636 A JP 2013-511076 A

Y. Boykov and G. Funka-Lea: “Interactive Graph Cuts for Optimal Boundary & Region Segmentation of Objects in ND Images”, Proceedings of “Interference Convence CV”. I, p. 105-112, July 2001.

  However, in the conventional technique using the graph cut, the outline of the head can often be accurately obtained due to the color difference between the head and the background, but the jaw outline is the same as the color of the jaw part. Since the color of the neck portion is similar, the contour obtained by the graph cut has a problem that it becomes an unnatural contour by rattling.

  An object of the present invention is to improve the accuracy of clipping of a specific area having contour portions having different color differences between the background, such as the contour of the head and the contour of the chin, for example, in the cutout of a face image.

In order to achieve the above object, an aspect of the image processing apparatus according to the present invention includes a specifying unit that specifies a specific region to be divided from the image in the image, and the boundary-likeness of the boundary of the specific region specified by the specifying unit. A boundary appropriate evaluation means for evaluating the appropriateness as the boundary based on the evaluation value, and an indefinite region that is in the vicinity of the boundary and includes the boundary based on the evaluation value A trimap creating means for creating a trimap comprising the specific region, a region other than the specific region, and the indefinite region for the image, and a graph cut method in the indefinite region of the trimap. and a dividing means for dividing a region other than the specific region and a specific region, the boundary diligence means further is specified by the specifying means In response to said color difference in the size of pixels in the specific region and the pixels other than the specific area, and changes the parameters relating to sensitivity for the evaluation.
In order to achieve the above object, according to one aspect of the image processing method of the present invention, a specific region to be divided from the image is specified in an image, and an evaluation value for evaluating the likelihood of the boundary of the specified specific region is calculated. And evaluating the adequacy as the boundary based on the evaluation value, setting a width of an indeterminate region that is in the vicinity of the boundary and includes the boundary based on the evaluation value, and specifying the specified for the image Create a trimap consisting of an area, an area other than the specific area, and the indefinite area, and divide the specific area and an area other than the specific area by the graph cut method in the indefinite area of the trimap as the boundary proper evaluation of further in response to said color difference in the size of pixels in the specific region and the pixels other than the specific region, to change the parameters related to the sensitivity for the evaluation Characterized in that it contains.
In order to achieve the above object, an aspect of the program of the present invention includes a step of identifying a specific area to be divided from the image in an image and calculating an evaluation value for evaluating the likelihood of the boundary of the identified specific area. And evaluating the appropriateness as the boundary based on the evaluation value, and setting a width of an indeterminate region that is near the boundary and includes the boundary based on the evaluation value, and for the image Creating a trimap comprising the specific region, a region other than the specific region, and the indefinite region, and dividing the specific region and the region other than the specific region in the indefinite region of the trimap by a graph cut method; Steps,
The cause the computer to execute Rutotomoni, the step of evaluating the appropriate, depending on the color difference in size between pixels other than the specific region and the pixel of the specific region specified in the step of the particular, for the evaluation The method includes a step of changing a parameter related to the sensitivity .
In order to achieve the above object, an aspect of the image processing apparatus according to the present invention includes a specifying unit that specifies a specific region to be divided from the image in the image, and the boundary-likeness of the boundary of the specific region specified by the specifying unit. A boundary appropriate evaluation means for evaluating the appropriateness as the boundary based on the evaluation value, and an indefinite region that is in the vicinity of the boundary and includes the boundary based on the evaluation value A trimap creating means for creating a trimap comprising the specific region, a region other than the specific region, and the indefinite region for the image, and a graph cut method in the indefinite region of the trimap. Dividing means for dividing a specific area and an area other than the specific area, wherein the image is a person image, and the specific area and the specific area The regions other than the region are a face image region and a background region, respectively, and the trimap creating means includes a boundary region between the chin and the neck in the face image region, and a head in the face image region other than the chin. The width of the indefinite region of the trimap is set to be different between the boundary region with the background region.
In order to achieve the above object, according to one aspect of the image processing method of the present invention, a specific region to be divided from the image is specified in an image, and an evaluation value for evaluating the likelihood of the boundary of the specified specific region is calculated. And evaluating the adequacy as the boundary based on the evaluation value, setting a width of an indeterminate region that is in the vicinity of the boundary and includes the boundary based on the evaluation value, and specifying the specified for the image Create a trimap consisting of an area, an area other than the specific area, and the indefinite area, and divide the specific area and an area other than the specific area by a graph cut method in the indefinite area of the trimap, and The image is a human image, and the specific area and the area other than the specific area are a face image area and a background area, respectively. The width of the indefinite region of the trimap is set to be different between the boundary region between the chin and the neck in the image region and the boundary region between the head and the background region in the face image region other than the chin. It is characterized by including.
In order to achieve the above object, an aspect of the program of the present invention includes a step of identifying a specific area to be divided from the image in an image and calculating an evaluation value for evaluating the likelihood of the boundary of the identified specific area. And evaluating the appropriateness as the boundary based on the evaluation value, and setting a width of an indeterminate region that is near the boundary and includes the boundary based on the evaluation value, and for the image Creating a trimap comprising the specific region, a region other than the specific region, and the indefinite region, and dividing the specific region and the region other than the specific region in the indefinite region of the trimap by a graph cut method; And causing the computer to execute the step, the image is a person image, and the specific region and a region other than the specific region Each of which is a face image area and a background area, and the step of creating the trimap includes a boundary area between a chin and a neck in the face image area, a head in the face image area other than the chin, and the background area. And a step of setting the width of the indefinite region of the trimap to be different from each other.

  According to the present invention, for example, it is possible to improve the accuracy of clipping of a specific region having a contour portion having a different color difference between the background such as the contour of the head and the contour of the chin in the clipping of the face image. .

It is a block diagram of an image processing device concerning this embodiment. It is a block diagram which shows the hardware structural example which implement | achieves the image processing apparatus which concerns on this embodiment. It is a flowchart which shows the example of the image processing which concerns on this embodiment. It is explanatory drawing of the determination operation | movement of the width | variety of an indefinite area | region. It is explanatory drawing of the clipping operation of the head in this embodiment. It is explanatory drawing of the clipping operation of the jaw in this embodiment. It is explanatory drawing of the cutting-out operation | movement of a face image.

  Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram of an image processing apparatus 100 according to an embodiment (hereinafter referred to as “this embodiment”) of the present invention. An image processing apparatus 100 according to the present embodiment includes a face detection unit 101, a contour acquisition unit 102, a trimap creation unit 103, a graph cut cutout unit 104, and a contour learning unit 105. The contour acquisition unit 102 further includes a head initial contour creation unit 106, a jaw initial contour creation unit 107, a head contour fitting unit 108, and a jaw contour fitting unit 109. The contour learning unit 105 further includes a head contour learning unit 110 and a jaw contour learning unit 111.

  The face detection unit 101 is a specifying unit that specifies a specific area to be divided in the input image. Specifically, the face detection unit 101 divides a face area and a background area in a person image.

  The contour acquisition unit 102 that operates as the boundary appropriateness evaluation unit calculates the evaluation value for evaluating the contour-likeness (boundary-likeness) of the contour (boundary) of the face area specified by the face detection unit 101, and calculates the evaluation value. Based on this, the appropriateness as a boundary is evaluated. In this embodiment, in order to acquire a contour, the contour learning unit 105 acquires a typical contour in advance by learning. More specifically, the head contour learning unit 110 acquires a representative head contour by learning, and the jaw contour learning unit 111 acquires a representative jaw contour by learning. Then, in the contour acquisition unit 102, the head initial contour creation unit 106 obtains the representative head contour acquired by the head contour learning unit 110 for the head of the face area specified by the face detection unit 101. An initial head contour is created by fitting. Similarly, the initial jaw contour creation unit 107 creates the initial jaw contour by applying the representative jaw contour acquired by the jaw contour learning unit 111 to the jaw of the face region. Furthermore, in the contour acquisition unit 102, the head contour fitting unit 108 deforms the head initial contour created by the head initial contour creation unit 106 while calculating and evaluating the evaluation value for the head contour. Yuki creates an optimal head contour. Similarly, while the jaw contour fitting unit 109 calculates and evaluates an evaluation value for the jaw contour, the initial jaw contour created by the initial jaw contour creation unit 107 is deformed to create an optimum jaw contour. .

Here, the head contour fitting unit 108 changes the parameter related to the sensitivity for evaluation in accordance with the magnitude of the color difference between the face area pixel and the background area pixel specified by the face detection unit 101. Can be configured. Specifically, the head contour fitting portion 108, the jaw and the neck of the boundary area can be configured differently the parameters related to the sensitivity for the boundary region, in assessment of the head and a background region other than the jaw .

The trimap creation unit 103 creates a trimap including a face region, a background region other than the face region, and an indeterminate region for the image while setting the width of the indefinite region near the contour based on the evaluation value. Specifically, the trimap creation unit 103 can set the width of the indefinite region of the trimap to be different for each different region according to the evaluation value described above. More specifically, the tri-map generating unit 103 sets the chin and the neck of the border area, and the boundary area between the head and the background region other than the chin, in so as to be different from the width of the unstable region of the tri map be able to.

  FIG. 2 is a block diagram illustrating a hardware configuration example that implements the image processing apparatus 100 of FIG. 1 according to the present embodiment.

  The image processing apparatus 100 is realized, for example, on a portable information terminal device such as a so-called smart phone terminal or a tablet terminal, a personal computer, or a computer system such as a digital signage.

  The image processing apparatus 100 includes a CPU (Central Processing Unit) 201, a ROM (Read Only Memory) 202, and a RAM (Random Access Memory) 203. The image processing apparatus 100 also includes an external storage device 204 such as a solid storage device or a hard disk storage device, a communication interface 205, an input device 206 such as a touch panel display device, and a display device 207. Further, the image processing apparatus 100 can set a portable recording medium 209 such as an SD / micro SD memory card, a USB (Universal Serial Bus) memory card, a compact flash memory card, a CD-ROM, a DVD, or a magneto-optical disk. A portable recording medium driving device 208 is provided. The imaging device 211 is a digital camera mechanism that can capture an image, and includes a lens, an autofocus drive control device, an exposure control device, an imaging sensor, and the like. In addition, you may make it acquire an image from the exterior, without providing the imaging device 211. FIG. The devices 201 to 208 and 211 described above are connected to each other by a bus 210.

  The ROM 202 stores a program for controlling general operations of the entire computer, as well as an image processing control program exemplified in the flowchart of FIG. The CPU 201 reads this control program from the ROM 202 and executes the RAM 203 as a work memory. Thereby, the image processing function shown by the functional block in FIG. 1 is realized. As a result, for example, image processing is performed in which a face image is cut out from a person image captured by the user with the imaging device 211. The face image data thus obtained is transferred to and used by various applications, for example, an application for creating a stamp or an application for displaying an electronic advertisement, and is displayed on the display device 207.

  The communication interface 205 is a device for connecting, for example, a LAN (local area network) or WAN (wide area network) communication line.

  The system according to the present embodiment is realized by the CPU 201 executing a program equipped with functions realized by the flowchart of FIG. The program may be distributed by being recorded in, for example, the external storage device 204 or the portable recording medium 209, or may be acquired from the network by the communication interface 205.

  Note that the image processing apparatus 100 according to the present embodiment may be realized on a server computer or the like instead of on a terminal device.

  FIG. 3 is a flowchart illustrating an example of image processing according to the present embodiment realized by the hardware configuration of the image processing apparatus 100 illustrated in FIG. Hereinafter, the hardware configuration of FIG. 2 will be referred to as needed.

  First, the CPU 101 executes face detection processing (step S301). The CPU 101 performs face detection in the original image to be recognized, and acquires the position and size of the face. For example, the CPU 101 learns feature quantities such as eyes, nose, and mouth of the distribution of facial brightness differences in the original image, and uses them to compare and detect. Alternatively, it may be performed by template matching or the like. This process realizes the function of the face detection unit 101 in FIG.

  Next, the CPU 101 calculates base data for the initial head contour and initial jaw contour (step S302). Here, for example, face detection and face head contour and chin contour detection are performed in advance from human images of about 200 samples. The detection of these contours is performed manually because an accurate contour is required. Then, for the position and size of the face acquired by face detection, for example, after normalizing the contour coordinates using the distance between the eye position and the interval, the average contour of all images is calculated, The calculation results are acquired as representative head contours and jaw contours. This process realizes the functions of the head contour learning unit 110 and the jaw contour learning unit 111 of FIG. Then, the CPU 101 determines the initial head contour CIh and the initial jaw contour CIa from the learning images of the head contour and the jaw contour thus obtained so as to match the face detected in step S301. . This process implements the functions of the initial head contour creation unit 106 and initial jaw contour creation unit 107 in FIG.

  Next, the CPU 101 sequentially deforms the initial head contour CIh, searches for an optimal contour while calculating an evaluation value for evaluating the deformed contour, and acquires the head contour Ch of the face image to be extracted. (Step S303). For example, the CPU 101 selects one pixel from the points constituting the initial contour CIh calculated in step S302, compares the vicinity of the head contour of the face image to be extracted with the pixel, and is suitable for the face contour. The evaluation value of whether or not the pixel is present is calculated. In this evaluation value calculation, for example, the CPU 101 adds 1 to the evaluation value when the pixel of the initial contour CIh is on the edge to be a contour in the original image, and adds −1 to the evaluation value when the pixel is not on the edge. . The CPU 101 repeats this process and selects the contour with the highest evaluation value. This process realizes the function of the head contour fitting unit 108 in FIG.

  Subsequently, as in the case of step S303, the CPU 101 sequentially deforms the initial jaw contour CIa, searches for an optimum contour while calculating an evaluation value for evaluating the deformed contour, and tries to cut out the jaw. The chin contour Ca of the face image is acquired (step S304). However, the evaluation value calculation method is changed between the case of the head contour Ch (step S303) and the case of the jaw contour Ca (step S304). In the case of the head, there is often a difference between the head and the background, and it is easy to appear as an edge. In the case of the chin, the difference between the chin and the neck is small and the edges are difficult to appear. Therefore, an edge image is created with a parameter (sensitivity higher than the head) so that many edges appear. For example, in the case of a jaw as an edge filter, detection is performed by weighting a lateral edge, or, for example, the threshold for detecting an edge with a differential value of luminance is lowered to increase edge extraction sensitivity. This process realizes the function of the jaw contour fitting unit 109 of FIG.

Next, CPU 101 from the acquired a head contour Ch and Agowa Guo Ca in step S303 and S304, to create a tri-map (step S305). The trimap is information used in the graph cut process in step S306, and is given to each pixel of the image as a foreground, background, or indefinite label. In this process, the CPU 101 makes the peripheral pixels of the head contour Ch and the jaw contour Ca acquired in steps S303 and S304 indefinite.

  At this time, as shown in FIG. 4, the CPU 101 increases the width Wh of the indefinite area for the peripheral pixels of the head outline Ch, and narrows the indefinite area for the peripheral pixels of the jaw outline Ca (for example, 1 pixel width). ) To avoid searching with graph cuts. As illustrated as (a) -1 or (a) -2 in FIG. 5, the head has individual hairstyle differences. In order to absorb the difference, as illustrated as (b) -1 or (b) -2 in FIG. 5, the indefinite area is made thicker and the indeterminate area is searched by a graph cut. In the case of the contour of the head, the contour can be obtained accurately due to the color difference between the head and the background, so that it can be thickened. As a result, as illustrated as (c) -1 or (c) -2 in FIG. 5, even if the hairstyle is different, a search is made in an indefinite region with a wide width Wh by a graph cut, so that the background of the head Can be divided appropriately.

  On the other hand, since there are few individual differences in the jaw, the indefinite region may be narrowed. Since the width of the indefinite region is narrow, the search for the graph cut is not performed so much. As a result, for example, as illustrated in FIG. 6A or 6B, the jaw can be smoothly divided from the background. If the indefinite region of the chin portion has the same width as the head, the graph cut performs a search in units of one pixel. Therefore, the color of the chin portion and the color of the neck portion are similar. As exemplified in c) or (d), the contour obtained is not smooth and unnatural.

As an example of determining the width Wh of the indefinite region from the evaluation value in step S305, the edge strength of the edge is high where the gradient strength is strong, and the width Wh of the indefinite region is reduced. Since the edge-like property is low at a place where the gradient strength is weak, a technique such as increasing the width Wh of the indefinite region can be adopted. The graph cut is a method of searching for an indefinite region obtained by the trimap and searching for an optimum boundary between the foreground and the background from the pixels. In the face clipping, an optimum face contour can be obtained by making the periphery of the face contour indefinite. At this time, the head contour Ch is the difference between the color of the hair and the background, and the contour can be accurately obtained by graph cutting, but the jaw contour Ca has almost no difference between the color of the chin and the neck. Therefore, it is difficult to obtain a correct contour by graph cut. Therefore, in this embodiment, in the creation of trimap, as shown in FIG. 4, a region having a width Wh around the head portion contour Ch and unstable region, the width of 1 pixel along the Agowa Guo Ca The area is an indefinite area.
Here, although there a region having a width of 1 pixel along the Agowa Guo Ca as unstable region, not limited to 1 pixel, may be two pixels or more.

  The above-described trimap processing in step S305 realizes the function of the trimap creating unit 103 in FIG.

  Thereafter, the CPU 101 executes graph cut processing using the trimap created in step S305 (step S306). Graph cut is a technique for dividing an image including a main object (in the case of the present embodiment, a face) into a main object area and a background area. When region division is performed, there may be a portion where the boundary is unclear due to the relationship between the main object and the background (hereinafter referred to as “indefinite region”), and it is necessary to perform optimal region division. Therefore, in this prior art, region division is regarded as an energy minimization problem, and a minimization method is proposed. In this prior art, the energy function is minimized by creating a graph that matches the region division and obtaining the minimum cut of the graph. This minimum cut realizes efficient area division calculation by using a maximum flow algorithm. As a specific calculation method, for example, the technique described in Non-Patent Document 1 can be employed. This process realizes the function of the graph cutout section 104 in FIG.

  Finally, the CPU 101 uses the result of the graph cut in step S306 to cut out the face area from the original image and output it (step S307).

  According to the embodiment described above, the contour of the head (head portion) of the trimap to be given to the graph cut is thickened so that a contour having a complicated shape can be acquired by the graph cut. In addition, by narrowing the jaw contour of the trimap given to the graph cut, the specified smooth and natural jaw contour can be acquired regardless of the graph cut. By calculating and acquiring the head contour and chin contour separately, fitting suitable for each contour can be performed, and the contour acquisition accuracy can be improved.

  In the present embodiment, the clipping of the face image has been described. However, the present invention is not limited to this. For example, a specific area other than a person such as a mascot character, an animal, or a flower may be clipped.

Regarding the above embodiment, the following additional notes are disclosed.
(Appendix 1)
A specifying means for specifying a specific area to be divided in the image;
Boundary appropriateness evaluation means for evaluating appropriateness as a boundary based on the evaluation value while calculating an evaluation value for evaluating the likelihood of the boundary of the specific area specified by the specifying means;
A trimap creating means for creating a trimap composed of the specific region, a region other than the specific region, and the undefined region for the image while setting a width of the undefined region near the boundary based on the evaluation value When,
A dividing unit that divides the specific region and a region other than the specific region by a graph cut method in an indefinite region of the trimap;
An image processing apparatus comprising:
(Appendix 2)
The image processing apparatus according to claim 1, wherein the specifying unit divides a face area and a background area in a person image.
(Appendix 3)
The boundary appropriate evaluation means changes the sensitivity-related parameters for the evaluation according to the color difference between the face area pixels and the background area pixels specified by the specifying means. 2. The image processing apparatus according to 2.
(Appendix 4)
The image processing according to appendix 3, wherein the boundary appropriateness evaluation means makes the parameter relating to the sensitivity for the evaluation different between a boundary region between a chin and a neck, and a head and a background region other than the chin. apparatus.
(Appendix 5)
The image according to any one of appendices 1 to 4, wherein the trimap creating means sets the width of the indefinite area of the trimap to be different for each different area according to the evaluation value. Processing equipment.
(Appendix 6)
The image according to appendix 5, wherein the tri-map creating means sets the width of the indefinite region of the tri-map to be different between the boundary region between the chin and the neck, and the head and background region other than the chin. Processing equipment.
(Appendix 7)
Identify specific areas to divide in the image,
While calculating the evaluation value that evaluates the likelihood of the boundary of the specific region specified by the specifying means, the suitability as a boundary is evaluated based on the evaluation value,
While setting the width of the indefinite area near the boundary based on the evaluation value, create a trimap consisting of the specific area, the area other than the specific area, and the indefinite area for the image,
Dividing the specific area and the area other than the specific area by a graph cut method in the indefinite area of the trimap,
An image processing method.
(Appendix 8)
Identifying a specific region to be divided in the image;
Evaluating the appropriateness as the boundary based on the evaluation value while calculating the evaluation value for evaluating the likelihood of the boundary of the boundary of the specific area specified by the specifying means;
Creating a trimap consisting of the specific region, a region other than the specific region, and the indefinite region for the image while setting a width of the indefinite region near the boundary based on the evaluation value;
Dividing the specific region and the region other than the specific region by a graph cut method in an indefinite region of the trimap;
A program that causes a computer to execute.

DESCRIPTION OF SYMBOLS 101 Face detection part 102 Contour acquisition part 103 Tri map creation part 104 Graph cut cutout part 105 Contour learning part 106 Head initial contour creation part 107 Initial jaw contour creation part 108 Head contour fitting part 109 Jaw contour fitting part 110 Head contour Learning unit 111 Jaw contour learning unit 111
201 CPU
202 ROM
203 RAM
204 External storage device 205 Communication interface 206 Input device 207 Output device 208 Portable recording medium driving device 209 Portable recording medium 210 Bus 211 Imaging device

Claims (11)

A specifying means for specifying a specific area to be divided from the image in the image;
A boundary appropriate evaluation means for calculating an evaluation value for evaluating the likelihood of the boundary of the boundary of the specific area specified by the specifying means, and evaluating the appropriateness as the boundary based on the evaluation value;
Based on the evaluation value, a width of an indefinite area that is in the vicinity of the boundary and includes the boundary is set, and a trimap that includes the specific area, an area other than the specific area, and the indefinite area for the image Tri-map creation means to create,
A dividing unit that divides the specific region and a region other than the specific region by a graph cut method in the indefinite region of the trimap;
Equipped with a,
The boundary appropriate evaluation means further changes a parameter relating to sensitivity for evaluation according to the magnitude of the color difference between the pixels of the specific area specified by the specifying means and the pixels other than the specific area. A featured image processing apparatus.   The image processing apparatus according to claim 1, wherein the specifying unit divides a face area as the specific area and a background area as an area other than the specific area from the person image in the person image. The boundary appropriate evaluation means is characterized in that the parameter relating to sensitivity for evaluation is different between a boundary region between a chin and a neck and a boundary region between a head other than the chin and the background region. The image processing apparatus according to claim 2. The tri-map creating means sets the width of the indefinite area of the tri-map to be different for each different area according to the evaluation value. the image processing apparatus. The trimap creating means sets the width of the indefinite region of the trimap to be different between a boundary region between the chin and the neck and a boundary region between the head and the background region other than the chin. The image processing apparatus according to claim 4 . The tri-map creating means sets the width of the indefinite region in the boundary region between the chin and the neck to be narrower than the width of the indefinite region in the boundary region between the head other than the chin and the background region. The image processing apparatus according to claim 5. Identifying a specific area to be divided from the image in the image;
Calculating an evaluation value for evaluating the likelihood of the boundary of the specified specific area, and evaluating the appropriateness as the boundary based on the evaluation value;
Based on the evaluation value, a width of an indefinite area that is in the vicinity of the boundary and includes the boundary is set, and a trimap that includes the specific area, an area other than the specific area, and the indefinite area for the image make,
Dividing the specific region and the region other than the specific region by the graph cut method in the indefinite region of the trimap,
The proper evaluation as the boundary further includes changing a parameter relating to the sensitivity for evaluation according to the magnitude of the color difference between the pixel in the specific region and the pixel other than the specific region. Image processing method. Identifying a specific region to be divided from the image in the image;
Calculating an evaluation value for evaluating the likelihood of the boundary of the specified specific area, and evaluating the appropriateness as the boundary based on the evaluation value;
Based on the evaluation value, a width of an indefinite area that is in the vicinity of the boundary and includes the boundary is set, and a trimap that includes the specific area, an area other than the specific area, and the indefinite area for the image A step to create,
Dividing the specific region and a region other than the specific region by a graph cut method in the indefinite region of the trimap;
To run on a computer,
The step of evaluating the appropriateness includes a step of changing a parameter relating to sensitivity for evaluation according to a magnitude of a color difference between the pixel in the specific region specified in the specifying step and a pixel other than the specific region. The program characterized by including. A specifying means for specifying a specific area to be divided from the image in the image;
A boundary appropriate evaluation means for calculating an evaluation value for evaluating the likelihood of the boundary of the boundary of the specific area specified by the specifying means, and evaluating the appropriateness as the boundary based on the evaluation value;
Based on the evaluation value, a width of an indefinite area that is in the vicinity of the boundary and includes the boundary is set, and a trimap that includes the specific area, an area other than the specific area, and the indefinite area for the image Tri-map creation means to create,
A dividing unit that divides the specific region and a region other than the specific region by a graph cut method in the indefinite region of the trimap;
With
The image is a person image, and the specific area and the areas other than the specific area are a face image area and a background area, respectively.
The trimap creating means includes the boundary region between the chin and the neck in the face image region, and the boundary region between the head and the background region in the face image region other than the chin, and the indefinite region of the trimap. The image processing apparatus is characterized in that the widths of the images are set differently. Identifying a specific area to be divided from the image in the image;
Calculating an evaluation value for evaluating the likelihood of the boundary of the specified specific area, and evaluating the appropriateness as the boundary based on the evaluation value;
Based on the evaluation value, a width of an indefinite area that is in the vicinity of the boundary and includes the boundary is set, and a trimap that includes the specific area, an area other than the specific area, and the indefinite area for the image make,
Dividing the specific region and the region other than the specific region by the graph cut method in the indefinite region of the trimap,
The image is a person image, and the specific area and the areas other than the specific area are a face image area and a background area, respectively.
In the creation of the trimap, the indeterminate of the trimap includes a boundary region between the chin and the neck in the face image region and a boundary region between the head and the background region in the face image region other than the chin. An image processing method comprising: setting different widths of regions. Identifying a specific region to be divided from the image in the image;
Calculating an evaluation value for evaluating the likelihood of the boundary of the specified specific area, and evaluating the appropriateness as the boundary based on the evaluation value;
Based on the evaluation value, a width of an indefinite area that is in the vicinity of the boundary and includes the boundary is set, and a trimap that includes the specific area, an area other than the specific area, and the indefinite area for the image A step to create,
Dividing the specific region and a region other than the specific region by a graph cut method in the indefinite region of the trimap;
To run on a computer,
The image is a person image, and the specific area and the areas other than the specific area are a face image area and a background area, respectively.
The step of creating the trimap includes a boundary region between a chin and a neck in the face image region, and a boundary region between the head and the background region in the face image region other than the chin. A program comprising the step of setting the width of the indefinite area to be different.