KR20110044392A - Image processing apparatus and method - Google Patents

Abstract

PURPOSE: An image processing apparatus and a method thereof are provided to separate an extracted interest region into sub regions and change the region color information of the sub region. CONSTITUTION: A silhouette extraction unit(110) extracts a silhouette region corresponding to a target object of a 3D model from an inputted color image. An interest region extraction unit(120) extracts a cloth region from the extracted silhouette region, and an interest region separation unit(130) separates the extracted interest region into plural sub regions. An image edition unit(140) substitutes some sub regions for other colors stored in a database(150).

Description

Image processing apparatus and method {IMAGE PROCESSING APPARATUS AND METHOD}

In a 3D model for a target object such as a human body of a user, it is associated with image processing for segmenting the 3D model, and more particularly, for image processing for changing texture information of the 3D model. .

Recently, with the development of image processing technology, interest in the field of 3D modeling of a target object such as a user's human body is also increasing. In 3D modeling of the human body, the human body may be used for virtual reality, computer graphic (CG) in a movie or video game.

The 3D model includes geometry information and texture information of the object. Here, the geometric information is represented by a plurality of meshes (or also called polygons) made of points or lines. The texture information is represented by texture information corresponding to each polygon, for example, a color value.

Meanwhile, the 3D model may be generated by the artist designing the geometric information and the texture information. In addition, in various industrial fields such as virtual reality, movies, and video games, a technology of generating a 3D model by directly scanning a target object, such as a human body, has also been studied.

In addition, a 3D model may be generated through image processing using a color image and a depth image of a human body.

However, in the case of 3D models generated by scanning a target object or image processing using a color image and a depth image of the target object, in addition to the 3D model designed and created by the artist, the generated geometric information and the texture information There may be some error in the matching of.

For example, the color value of the upper part matches the polygon of the lower part, and so on. Therefore, when the region of interest is separated from the 3D model or separated into a plurality of regions, the utilization is expected to be high.

Provided are an image processing apparatus and method for extracting a specific region of interest from a target object such as a 3D model of a human body.

There is also provided an image processing apparatus and method for dividing an extracted region of interest into several small regions and changing color information of some of the small regions.

According to an aspect of the present invention, a silhouette extractor for extracting a silhouette region corresponding to a target object in the 3D model from an input color image matched with an input 3D model, a region of interest extractor for extracting a region of interest from the silhouette region; And a region of interest separator configured to separate the extracted region of interest into a plurality of sub regions.

Wherein the target object is, for example, a human body.

The silhouette extractor may include: a first calculator configured to project a human body part of the input 3D model to the input color image, and select a first pixel group corresponding to the projected part; Each of the pixels is classified into a pixel of a first type belonging to the target object, a pixel of a second type not belonging to the target object, and a pixel of a third type not classified as either the first type or the second type. Obtaining a second color probability distribution function for the pixels of the first type and a second color probability distribution function for the pixels of the second type; If the result of substituting the color value of one pixel into the first color probability distribution function is greater than the result of substituting the color value of the first pixel into the second color probability distribution function, the first pixel is assigned to the first prevalence. It may comprise a third calculation unit configured to determine a pixel-by-pixel.

Here, the first color probability distribution function and the second color probability distribution function may be Gaussian functions.

According to some embodiments of the invention, the region of interest is a cloth region. In this case, the ROI extractor may include a first type of pixel belonging to the clothing area of the human body, a second type of pixel not belonging to the clothing area, and each of the plurality of pixels included in the silhouette area. A first calculation unit classifying a pixel of a third type not classified into one of the types or the second type, and a first color probability distribution function for the pixels of the first type and the pixels of the second type. Obtaining a second color probability distribution function for the second color probability and substituting the color value of the first pixel of the third type of pixels into the first color probability distribution function If greater than the result of substituting the distribution function, a second calculator may be configured to determine the first pixel as the first type pixel.

In this case, the second type of pixels are pixels corresponding to at least one of a face, a hand, and a foot of a human body.

According to an embodiment of the present invention, the plurality of sub regions includes at least one of a lower region, an upper shirt region, and an upper jacket region.

The image processing apparatus may further include a database configured to store image data corresponding to at least one of the plurality of separated small regions, and a first small region selected from the plurality of separated small regions using data stored in the database. The apparatus may further include an image editor configured to change an image of the area.

According to another aspect of the present invention, extracting a silhouette region corresponding to a target object in the 3D model from the input color image matched with the input 3D model, extracting the region of interest from the silhouette region, and the extracted An image processing method is provided, comprising separating a region of interest into a plurality of small regions.

According to the above embodiments, an application for changing color information of a part of the 3D model by image processing of extracting a specific region of interest, such as a clothes region, from the 3D model of the human body and separating the region of interest into a small region such as a top and a bottom, etc. This is possible.

According to the image processing, it is possible to change the clothes of the avatar in the virtual reality, the character in the movie or the video game, or to simulate the user's avatar with the item to be purchased in the online shopping in advance.

Hereinafter, some embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited or limited by the embodiments. Like reference numerals in the drawings denote like elements.

1 illustrates an image processing apparatus 100 according to an embodiment of the present invention.

In the image processing apparatus 100 of the present invention, geometric information and texture information of a color image and a 3D model matched with the color image may be input.

The silhouette extractor 110 extracts a target object of the 3D model, for example, a silhouette region corresponding to a human body, from the input color image.

The operation of the silhouette extractor will be described later in more detail with reference to FIGS. 4 to 6.

However, when the 3D model is generated by an image processing method using the color image, a silhouette region corresponding to the target object of the 3D model may have already been extracted. The image processing apparatus and the image processing method according to the embodiment of the present invention do not exclude such a possibility, and in this case, the silhouette extracting unit 110 may bring the silhouette region extracted from the color image as it is, or slightly modify it. You can get it by adding

The ROI extractor 120 extracts an ROI, for example, a clothes region, from the extracted silhouette region. An operation of the ROI extractor 120 will be described later in more detail with reference to FIGS. 6 to 9.

The ROI separator 130 divides the extracted ROI into a plurality of subregions.

For example, when the extracted region of interest is a clothes region, the region of interest separator 130 may divide it into a jacket portion small region, a shirt portion small region and a lower portion small region.

The operation of the ROI separator 130 will be described later in more detail with reference to FIGS. 9 through 10.

When the ROI is divided into a plurality of small regions, the image editing unit 140 may replace at least a portion of the ROI with another image previously stored in the database 150.

For example, the jacket collar may be changed by changing the jacket portion small area.

The operation of the image editing unit 140 will be described later in more detail with reference to FIG. 11.

Hereinafter, the operation of each configuration will be described through specific embodiments.

2 illustrates a color image 200 input according to an embodiment of the present invention.

In the present embodiment, the color image 200 is matched with the 3D model shown in FIG. 3.

The color image 200 includes a human body part 210, a back background part 201, and a bottom part 202.

According to the image processing method according to the exemplary embodiment of the present invention, the silhouette region corresponding to the human body 210 is extracted from the color image 200 by the silhouette extractor 110 of FIG. 1. In addition, a region of interest (ROI) is extracted from the extracted silhouette region by the region of interest extractor 120.

The region of interest extracting unit 120 removes the face portion 251, the hand portion 252, the shoe portion 253, and the like from the silhouette region, and extracts the outer garment region as the region of interest.

The clothes region, which is the region of interest to be extracted, includes the upper jacket 220 small region, the upper shirt small region 240, and the lower small region 230. In this case, each of the small regions may be separated by the ROI separator 130.

When the small areas are separated in this way, some of the small areas separated by the image editing unit 140 may be changed using other data stored in the database 150.

Details of each step of the image processing, such as silhouette region extraction, region of interest extraction, region of interest separation, and the like, will be described below with reference to FIG. 3.

3 illustrates a 3D model 300 input according to an embodiment of the present invention.

The 3D model includes a target object 310. In the present embodiment, the target object 310 corresponds to a human body.

As described above, the 3D model is represented by geometric information and texture information, and the polygon 311 is an example included in the geometric information of the target object 310.

In the following embodiments, the 3D model 300 is not designed by an artist, but is scanned by a target object or generated by image processing using a color image and a depth image of the target object. It is assumed that Of course, this does not exclude the case that the 3D model is generated by some other method for modeling an object in the real world rather than a 3D model designed directly.

For image processing according to an embodiment of the present invention, the color image 200 of FIG. 2 and the 3D model 300 of FIG. 3 are input.

Here, the color image 200 and the 3D model 300 are matched with each other. Thus, information about which pixels in the color image 200 each polygon of the 3D model, such as polygon 311, is given together.

For example, when the 3D model 300 is generated by an image processing method using a color image 200 and a depth image (not shown) that matches the color image 200, the 3D model 300 may be configured. It may be understood that information of which pixels in each of the polygons corresponds to the color image 200 is also given in advance.

According to an image processing method according to an embodiment of the present invention, each polygon of the target object 310 of the 3D model, for example, the polygon 311 is projected onto the color image 200.

The projection is performed by a first calculator (not shown) of the silhouette extractor 110 of FIG. 1.

Then, pixels corresponding to the target object 310 may be selected by using which pixel in the color image 200 each of the polygons is projected.

4 illustrates a result of projecting the target object 310 of the 3D model of FIG. 3 onto the input color image 200 of FIG. 2, according to an exemplary embodiment.

The locations where the respective polygons of the target object 310 are projected on the color image 400 are indicated by scattered points.

However, although the target object 310 of the 3D model 300 and the human body 210 of the color image 200 are matched, there may be an error due to a matching error.

In this embodiment, some errors are found, such as some of the points being located in the back background portion 201 of the color image 200.

If there is no matching error, the point of the outermost part of the point will be located exactly at the boundary of the human body part 210, such as point 410. However, due to the above matching error, there are also points located in the background part 201, such as the point 420, and also points located inward of the boundary of the human body 210, such as the point 430. do.

Therefore, when the pixels in which the points are located are directly extracted into the silhouette region of the human body 210, there is a possibility that the rear background portion 201 may be included together.

Therefore, the silhouette extractor 110 extracts only a portion of the pixels in which the points are located as the silhouette region corresponding to the human body 210 through an additional process.

This process is described with reference to FIG.

FIG. 5 illustrates a result of extracting a projected part of the 3D model from the input color image 200 of FIG. 4 and dividing the extracted part into three areas according to an embodiment of the present invention.

Region 510 is an area that is considered to correspond to the human body portion 210. Even if the above matching error exists, the degree of the error will not be seriously large, and therefore, among some pixels selected by the first calculation unit of the silhouette extraction unit 110 as a result of the projection, it is equal to or greater than a predetermined first threshold value. Pixels located inward by the distance are considered to correspond to the human body portion 210.

In addition, the area 520 is an area deemed not to correspond to the human body part 210. Although the outer point may accurately reflect the boundary of the human body 210 as in the point 410 of FIG. 4, the outer point may be located at the rear background part 201 as in the point 420. Therefore, to correct this matching error, the area 520 is considered to not correspond to the human body portion 210.

In this case, pixels existing at a distance less than a predetermined second threshold value from the edge may be determined as the area 520.

The remaining pixels determined to not belong to either the region 510 or the region 520 are determined as the region 530.

This area 530 is an ambiguous area that is not considered to correspond to the human body portion 210 certainly, and is not necessarily considered to correspond to the human body portion 210.

Separation of the regions 510 to 530 is performed by the second calculator of the silhouette extractor 110. The first threshold and the second threshold may be preset values, or may be set or adjusted by a user.

In embodiments of the present disclosure, the process of determining again the pixels belonging to the region 530 as belonging to the region 510 or the region 520 is performed by the third calculator of the silhouette extractor 110. do.

Here, the third calculator of the silhouette extractor 110 calculates a probability distribution function of color values with respect to color values of all pixels belonging to the region 510.

For example, the first color probability distribution function Pr (c) for all pixels belonging to the region 510 may be expressed as a linear combination of K Gaussian distributions.

[Equation 1]

는 양수이고,

이다.here

Is positive,

to be.

C is a vector having three color elements of R (Red), G (Green), and B (Blue), and C = (R, G, B). Of course, the color coordination system is not an RGB system, but a YCbCr system, x.v. In the case of a color system or the like, the C may be a vector of another dimension.

는 Gaussian 분포이고,

는 Gaussian 분포의 평균,

는 Gaussian 분포의 Covariance matrix 이다.Also in accordance with one embodiment of the present invention,

Is the Gaussian distribution,

Is the mean of the Gaussian distribution,

Is the covariance matrix of the Gaussian distribution.

In addition, Pr (c) may be obtained by using an Expectation Maximization technique or a Principal Component Analysis (PCA) technique. Such a technique is merely an example known to those skilled in the art.

The second color probability distribution function Pb (c) for all pixels belonging to the region 520 may also be expressed as a linear combination of K Gaussian distributions.

Then, when the first color probability distribution function Pr (c) and the second color probability distribution function Pb (c) are obtained, the third calculation unit of the silhouette extractor 110 may apply to each of all pixels belonging to the region 530. For example, the value substituted into the first color probability distribution function and the value substituted into the second color probability distribution function are compared.

For example, among all the pixels belonging to the region 530, Pr (Cn), which is a result of substituting the color vector Cn of the nth pixel into the first color probability distribution function, and the second color probability distribution function, are substituted. Compare the resultant Pb (Cn).

If Pr (Cn) is larger than Pb (Cn), the n-th pixel is determined to belong to the region 510. Otherwise, the nth pixel is determined to belong to the region 520.

By performing this process on each of all the pixels belonging to the area 530, the final silhouette area belonging to the human body part 210 is now extracted.

Meanwhile, according to an embodiment of the present invention, in the above process, an optimization method (Graph Cuts, Belief Propagation) based on Markov Random Field Modeling, which is known to those skilled in the art, may be applied.

FIG. 6 illustrates a result 600 of extracting the silhouette region 610 using the result of FIG. 5, according to an exemplary embodiment.

The extracted silhouette area 610 includes pixels corresponding to the human body part 210.

As described above, in the present exemplary embodiment, among the silhouette regions 610 extracted by the ROI extractor 120 of FIG. 1, the face region 621, the hand region 622, and the shoe region 623 may be formed. The clothes area 611, 612, etc. are extracted as the region of interest ROI.

A process of extracting the clothes region, which is the region of interest, from the silhouette region 610 by the region of interest extractor 120 will be described with reference to FIG. 7.

FIG. 7 illustrates the result 700 of dividing the silhouette region 610 of FIG. 6 into three portions, according to one embodiment of the invention.

According to one embodiment of the invention, the first calculation unit (not shown) of the region of interest extraction unit 120, the silhouette area 610, the area 710 that is considered to belong to the clothes area, clothes area It is divided into the region 720 and the remaining region 730 which are not considered to belong to any of the regions 720 and 710.

This process is similar to the operation of the silhouette extraction unit 110.

For example, the first calculation unit of the region of interest extractor 120 determines that the head having the training data belongs to the region 720, and the pixel is sure to belong to the clothes portion (or the body portion). Determine that they belong to the area 710.

In the meantime, a band having a predetermined width is determined as the region 730 where the judgment is uncertain.

Then, the second calculator (not shown) of the ROI extractor 120 obtains a third color probability distribution function for all the pixels in the area 710, and all the pixels in the area 720. Obtain a fourth color probability distribution function for.

For each of the pixels in the area 730, the result of substituting the third color probability distribution function with the result of substituting the fourth color probability distribution function is compared with each other. Each of the pixels is again determined to belong to either the region 710 or the region 720.

By this process, the head (or face) area of the human body is removed from the silhouette area 610.

FIG. 8 illustrates a result 800 of removing a face region other than the region of interest from the silhouette region by using the result of FIG. 7 according to an embodiment of the present invention.

In this case, when the face region is removed, the hand region 841 and the shoe region 842 are present as they are, in addition to the clothes regions 810, 820, and 830.

In this case, the ROI extractor 120 repeatedly performs the process of FIG. 7 to remove the hand area 841 and the shoe area 842, and the clothes area 810, 820, and the ROI. 830).

FIG. 9 illustrates a result 900 of extracting a clothes region 910 of a region of interest from the silhouette region 610 according to the method of FIG. 8 according to an exemplary embodiment of the present invention.

Meanwhile, according to an exemplary embodiment, the ROI separator 130 of FIG. 1 separates the clothes region 910 thus extracted into a plurality of subregions 1010 to 1030.

In this case, the first calculation unit (not shown) of the region of interest separation unit 130 divides the clothes region 910 into a region that does not certainly belong to the region that belongs to the lower portion. In this case, the training data can be utilized.

The second calculator (not shown) of the ROI separator 130 obtains a probability distribution function similar to the above Equation 1 for each of the regions.

The second calculation unit of the ROI separator 130 compares the result of substituting each of the obtained probability distribution functions with respect to each pixel of the boundary portion which does not belong to any of the areas. It can be determined whether the pixel belongs to the lower part or not, and thus the lower part small area and the other part can be separated.

The detailed process may be understood similarly to that described above with reference to FIG. 5 and Equation 1.

Furthermore, the ROI separator 130 may separate the jacket portion small region and the shirt portion small region through a similar process.

FIG. 10 illustrates a result 1000 of dividing the region of interest 910 extracted from FIG. 9 into a plurality of small regions according to an embodiment of the present invention.

Small region 1010 corresponds to the shirt portion of the top, small region 1020 corresponds to the jacket portion of the top, and small region 1030 corresponds to the bottom portion.

Through this process, clothes of the human body are separated by type in the input color image 200 relatively accurately and simply.

In addition, since the target object 310 of the 3D model 300 of FIG. 3 is matched with the input color image 200, which of the polygons of the target object 310 corresponds to the top shirt, and which top It may be determined which corresponds to the jacket and which corresponds to the bottoms.

According to an embodiment of the present invention, at least some of the divided small regions 1010, 1020, and 1030 may be replaced with other image data stored in the database 150 of FIG. 1. In other words, you can change your jacket to something else, or your pants to something else.

This image editing process is performed by the image editing unit 140 of FIG. The image editing unit 140 may change some pixel values of the color image 200 or may change texture information of some polygons of the target object 310 of the 3D model.

FIG. 11 illustrates a result 1100 of changing a portion of the plurality of small regions of FIG. 10 to another image according to an embodiment of the present invention.

The small region 1010 corresponding to the shirt was replaced by a small region 1110, the small region 1020 corresponding to the jacket was replaced by a small region 1120, and the small region 1030 corresponding to the bottom was a small region. (1130).

According to another embodiment of the present invention, the region of interest separator 140 may further separate the small region 1111 corresponding to the tie portion, and in this case, image processing for changing the tie to another one is also possible.

According to the image processing apparatus and the image processing method described through the above process, in the virtual reality field, the clothes of the 3D model avatar generated based on the color image of the user may be changed to another.

It is also possible to change the clothes of characters in a movie or video game modeling a specific person.

Furthermore, in the field of online shopping, which is currently increasing in interest, it is possible to perform a simulation of pre-coating a jacket to be purchased on a virtual model identical to a user's body shape, which can be a great opportunity for the growth of the online industry.

12 illustrates an image processing method according to an embodiment of the present invention.

In operation S1210, a target object of a 3D model, for example, a silhouette region corresponding to a human body is extracted from the input color image 200. Extraction of the silhouette region is as described above with reference to FIGS. 4 to 6.

Then, the region of interest, such as the clothes region, is extracted from the silhouette region extracted in step S1220.

The process of extracting the clothes area, that is, the clothes area 910 of FIG. 9, from the silhouette area 610 of FIG. 6 is as described above with reference to FIGS. 6 to 9.

In operation S1230, the extracted region of interest is divided into a plurality of small regions. The above-described region of interest separation process described above with reference to FIGS. 9 to 10 is merely an embodiment of the present disclosure, and the type of the region of interest and the number or type of the small regions that divide the region of interest may be changed. . It will be apparent to those skilled in the art that such changes in degree cannot be interpreted as departing from the spirit of the present invention.

In operation S1240, an image of some of the separated small regions may be replaced with another image stored in a database. This process is as described above with reference to FIG. In this case, a part of the small region may be changed for the input color image 200, and also for texture information of the target object 310 of the 3D model 300.

If the user renders a 3D model in which the texture information of the small area is changed, as in the above-described application example, it is possible to simulate a specific clothes on the avatar in the virtual reality.

Method according to an embodiment of the present invention is implemented in the form of program instructions that can be executed by various computer means may be recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from these descriptions. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined not only by the claims below but also by the equivalents of the claims.

1 illustrates an image processing apparatus according to an embodiment of the present invention.

2 illustrates a color image input according to an embodiment of the present invention.

3 illustrates an input 3D model according to an embodiment of the present invention.

FIG. 4 illustrates a result of projecting the target object of the 3D model of FIG. 3 onto the input color image of FIG. 2, according to an exemplary embodiment.

FIG. 5 illustrates a result of extracting a portion from which the 3D model is projected from the input color image of FIG. 4 and separating the extracted portion into three regions again according to an embodiment of the present invention.

FIG. 6 illustrates a result of extracting a silhouette region using the result of FIG. 5, according to an exemplary embodiment.

FIG. 7 illustrates a result of dividing the silhouette region of FIG. 6 into three parts, according to an embodiment of the present disclosure.

FIG. 8 illustrates a result of removing a face region other than the region of interest from the silhouette region using the result of FIG. 7 according to an embodiment of the present invention.

FIG. 9 illustrates a result of extracting a clothes region, which is a region of interest, from the silhouette region by the method of FIG. 8 according to an embodiment of the present invention.

FIG. 10 illustrates a result of dividing the region of interest extracted in FIG. 9 into a plurality of small regions according to an embodiment of the present invention.

FIG. 11 illustrates a result of changing a portion of the plurality of small regions of FIG. 10 to another image according to an embodiment of the present invention.

12 illustrates an image processing method according to an embodiment of the present invention.

Claims (18)

A silhouette extractor configured to extract a silhouette region corresponding to a target object in the 3D model from an input color image matched with an input 3D model; An ROI extractor configured to extract an ROI from the silhouette region; And A region of interest separator that separates the extracted region of interest into a plurality of sub regions. Image processing apparatus comprising a. The method of claim 1, The target object is a human body (human body), the image processing device. The method of claim 2, The silhouette extraction unit, A first calculator configured to project a human body part of the input 3D model onto the input color image to select a first pixel group corresponding to the projected part; Each of the pixels of the first pixel group is classified into a first type of pixel belonging to the target object, a second type of pixel not belonging to the target object, and a first type or a second type not classified into the second type. A second calculation unit classifying the three types of pixels; And A color probability distribution of the first type of pixels and a color probability distribution of the second type of pixels are obtained, and with reference to the obtained color probability distributions, each of the third type of pixels is referred to as the first type or the second type. Third calculation unit judged to belong to Image processing apparatus comprising a. The method of claim 3, The third calculator calculates a first color probability distribution function for the first type of pixels and a second color probability distribution function for the second type of pixels, and obtains a first one of the third type of pixels. If the result of substituting the color value of the pixel into the first color probability distribution function is greater than the result of substituting the color value of the first pixel into the second color probability distribution function, the first pixel is converted into the first type pixel. Determined by, the image processing device. The method of claim 4, wherein And the first color probability distribution function and the second color probability distribution function are Gaussian functions. The method of claim 2, And the region of interest is a cloth region. The method of claim 6, The region of interest extraction unit, Each of the plurality of pixels included in the silhouette area is one of a first type of pixel belonging to the clothing area of the human body, a second type of pixel not belonging to the clothes area, and any one of the first type or the second type. A first calculation unit classifying the third type of pixels which are not classified; And A color probability distribution of the first type of pixels and a color probability distribution of the second type of pixels are obtained, and with reference to the obtained color probability distributions, each of the third type of pixels is referred to as the first type or the second type. Second calculation unit judged to belong to Image processing apparatus comprising a. The method of claim 7, wherein The second calculator calculates a first color probability distribution function for the first type of pixels and a second color probability distribution function for the second type of pixels, and obtains a first one of the third type of pixels. If the result of substituting the color value of the pixel into the first color probability distribution function is greater than the result of substituting the color value of the first pixel into the second color probability distribution function, the first pixel is converted into the first type pixel. Determined by, the image processing device. The method of claim 7, wherein The pixel of the second type is a pixel corresponding to at least one of a face, a hand, and a foot of a human body. The method of claim 6, The plurality of sub regions may include at least one of a lower region, an upper shirt region, and an upper jacket region. The method of claim 10, The region of interest separator, Each of the plurality of pixels included in the clothes area is not classified into a first type of pixel belonging to the lower area, a second type of pixel not belonging to the lower area, and one of the first type or the second type. A first calculating unit to classify the non-third type pixel; And Obtaining a first color probability distribution function for the first type of pixels and a second color probability distribution function for the second type of pixels, and calculating the color value of the first one of the third type of pixels; A second calculation for determining the first pixel as the first type pixel when the result of substituting the first color probability distribution function is greater than the result of substituting the color value of the first pixel with the second color probability distribution function; part Image processing apparatus comprising a. The method of claim 11, The region of interest separator, Each of the second type of pixels is classified into a fourth type of pixel belonging to the shirt area, a fifth type of pixel not belonging to the shirt area, and a sixth type not classified into any of the fourth type or the fifth type. A third calculation unit classifying the pixel of the type; And Obtaining a third color probability distribution function for the fourth type of pixels and a fourth color probability distribution function for the fifth type of pixels, and calculating a color value of a second pixel of the sixth type of pixels; A fourth calculation that determines the second pixel as the fourth type pixel when the result of substituting the third color probability distribution function is greater than the result of substituting the color value of the second pixel into the fourth color probability distribution function part Further comprising, the image processing device. The method of claim 1, A database storing image data corresponding to at least one of the plurality of separated small regions; And An image editing unit configured to change an image of a first selected small area among the plurality of separated small areas by using data stored in the database Further comprising, the image processing device. Extracting a silhouette region corresponding to a target object in the 3D model from an input color image matched with the input 3D model; Extracting a region of interest from the silhouette region; And Separating the extracted region of interest into a plurality of small regions Image processing method comprising a. The method of claim 14, Extracting the silhouette region, Projecting a human body part of the input 3D model onto the input color image to select a first pixel group corresponding to the projected part; Each of the pixels of the first pixel group is classified into a first type of pixel belonging to the target object, a second type of pixel not belonging to the target object, and a first type or a second type not classified into the second type. Classifying into three types of pixels; And Obtaining a first color probability distribution function for the first type of pixels and a second color probability distribution function for the second type of pixels, and calculating the color value of the first one of the third type of pixels; Determining the first pixel as the first type pixel when the result of substituting the first color probability distribution function is greater than the result of substituting the color value of the first pixel with the second color probability distribution function. Image processing method comprising a. The method of claim 14, Extracting the ROI may include: Each of the plurality of pixels included in the silhouette region is not classified into a first type of pixel belonging to the ROI, a second type of pixel not belonging to the ROI, and one of the first type or the second type. Categorizing into a third type of pixel; And Obtaining a first color probability distribution function for the first type of pixels and a second color probability distribution function for the second type of pixels, and calculating the color value of the first one of the third type of pixels; Determining the first pixel as the first type pixel when the result of substituting the first color probability distribution function is greater than the result of substituting the color value of the first pixel with the second color probability distribution function. Image processing method comprising a. The method of claim 14, Reading image data corresponding to a first selected small area of the separated plurality of small areas from a database; And Changing the image of the first small region by using the read data; Further comprising, the image processing method. A computer-readable recording medium containing a program for performing the image processing method of any one of claims 14 to 17.

Images