KR20130081080A - Apparatus and method for color image boundary clearness - Google Patents
Apparatus and method for color image boundary clearness Download PDFInfo
- Publication number
- KR20130081080A KR20130081080A KR1020120002051A KR20120002051A KR20130081080A KR 20130081080 A KR20130081080 A KR 20130081080A KR 1020120002051 A KR1020120002051 A KR 1020120002051A KR 20120002051 A KR20120002051 A KR 20120002051A KR 20130081080 A KR20130081080 A KR 20130081080A
- Authority
- KR
- South Korea
- Prior art keywords
- image
- boundary
- pixel
- value
- weight
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/46—Colour picture communication systems
- H04N1/56—Processing of colour picture signals
- H04N1/58—Edge or detail enhancement; Noise or error suppression, e.g. colour misregistration correction
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/222—Studio circuitry; Studio devices; Studio equipment
- H04N5/2224—Studio circuitry; Studio devices; Studio equipment related to virtual studio applications
- H04N5/2226—Determination of depth image, e.g. for foreground/background separation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20076—Probabilistic image processing
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
Description
The present invention relates to an image boundary clearing device and method and a recording medium therefor. More specifically, the present invention calculates weights by using a random walk probability model in a bilateral filter and averages the adjacent pixels using the calculated weights to preserve the information on the image boundary while clearing the boundary. The present invention relates to an apparatus and method for clarifying image boundaries and a recording medium therefor.
Recently, as information processing technology for broadcasting and communication is rapidly developed, interest in the next generation broadcasting service is increasing. In this regard, studies on stereoscopic video and multi-view video (MVC) have been conducted to provide realistic three-dimensional stereoscopic images that are differentiated from conventional two-dimensional images. In addition, research on free-viewpoint TV (FTV), which provides an image at any point desired by a user, is also in progress.
In these techniques, a depth image may be effectively used when synthesizing an image, generating a multiview image, or generating an image for an intermediate view. Unlike a color image, a depth image has depth information and contour information of an object, not texture information.
Bilateral filters are used for noise removal and planarization while preserving the boundary of an image. In the case of a general two-sided filter, weights based on color and distance are calculated. The values of adjacent pixels are averaged using the calculated weights. In general, both filters use a method of preserving boundary information by largely calculating weights between close and similar colors.
In the case of a general two-sided filter, there is a problem in that boundaries are not clear for an ambiguous color gamut.
As another bilateral filter method, a boundary between an image and a depth map may be matched to clarify the boundary of the image.
When using this method, the exact depth value actually captured and acquired may be caused by the weighted sum of the surrounding prediction values, and the accuracy of the boundary may not be accurate in the obscure color region around the boundary due to the influence of the initial depth information. There is a problem falling.
The present invention has been made to solve the above problems, and the weight is calculated using the random walk probability, and the reliability of the depth map and the boundary of the image are clearly defined using the calculated weight. It is an object of the present invention to provide an apparatus and method for clarifying image boundaries and a recording medium therefor.
In order to solve the above technical problem, the image boundary disambiguation method according to the present invention is a method for clarifying boundary portions of an image using a bilateral filter, using weights using random walk probability. And a random walk weight calculation step of calculating.
Preferably, the method may further include a filtering step of averaging and filtering values of pixels of an adjacent image by using the weight.
Preferably, the weight using the random walk probability is a value obtained by subtracting the sum of cost values from the pixel at the boundary portion to the pixel at the center portion in the initial probability 1.
Preferably, the cost value is a value obtained by calculating a color difference using a Gaussian distribution between adjacent pixels.
Preferably, the method further includes a block classification step of dividing the entire area of the image into block units having a predetermined size.
The weight calculation may be calculated in units of the divided blocks.
Preferably, the sum of the cost values corresponds to one of a path having the maximum weight or a path having the lowest cost value when two or more paths from the boundary pixel to the center pixel exist. It is characterized in that the selected value calculated.
Preferably, the path having the maximum random walk probability or the path having the smallest cost value is calculated by applying the Dirichlet Problem method.
Preferably, the filtering is performed on a depth map of the image.
In order to solve the above technical problem, the image boundary disambiguation apparatus according to the present invention is a device for clarifying boundary portions of an image using a bilateral filter. The weight is determined using random walk probability. It characterized in that it comprises a random walk weight calculation unit to calculate.
The apparatus may further include a filtering unit which averages and filters values of pixels of an adjacent image by using the weight.
Preferably, the weight using the random walk probability is a value obtained by subtracting the sum of cost values from the pixel at the boundary portion to the pixel at the center portion at the initial probability 1.
Preferably, the cost value is a value obtained by calculating a color difference using a Gaussian distribution between adjacent pixels.
Preferably, the apparatus further includes a block separator for dividing the entire area of the image into block units having a predetermined size, and the weight calculation is performed in the divided block units.
Preferably, the sum of the cost values corresponds to one of a path having the maximum weight or a path having the lowest cost value when two or more paths from the boundary pixel to the center pixel exist. It is characterized in that the selected value calculated.
Preferably, the calculation of the path having the maximum random walk probability or the path having the smallest cost value is calculated by applying a Dirichlet Problem apparatus.
Preferably, the filtering is performed on a depth map of the image.
According to the present invention, weights are calculated using random walk probabilities, and applied to both filters using the calculated weights, thereby improving reliability of a depth map and clarifying image boundaries.
In addition, the present invention can reduce the problem of reducing the overall resolution of the image by generating chromatic aberration of the image.
1 is a block diagram of an image boundary disambiguation apparatus according to a preferred embodiment of the present invention.
FIG. 2 is a diagram for explaining a center pixel and a boundary pixel in an entire area of an image when a weight is calculated in units of blocks according to the present invention.
3 is an exemplary diagram for explaining a cost value in an image boundary disambiguation apparatus according to the present invention.
4 is an exemplary diagram illustrating a process of calculating a weight in an image boundary disambiguation apparatus according to the present invention.
5 is an exemplary view for explaining a method of calculating a cost value when a path is 2 or more in the image boundary disambiguation apparatus according to the present invention.
6 is another exemplary view for explaining a method of calculating a cost value when a path is 2 or more in the image boundary disambiguation apparatus according to the present invention.
7 is a flowchart of an image boundary method according to a preferred embodiment of the present invention.
8 is an exemplary diagram for describing an example in which an image boundary disambiguation method according to the present invention is applied to a depth map.
9 is an exemplary diagram for explaining another example of applying the image boundary disambiguation method according to the present invention to a depth map.
Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. In the following description and the accompanying drawings, substantially the same components are denoted by the same reference numerals, and redundant description will be omitted. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.
It is to be understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, . On the other hand, when an element is referred to as being "directly connected" or "directly connected" to another element, it should be understood that there are no other elements in between.
1 is a block diagram of an image boundary disambiguation apparatus according to a preferred embodiment of the present invention.
Referring to FIG. 1, the image boundary disambiguation apparatus 100 includes a random walk weight calculator 120, a block divider 110, and a filter 130.
The random walk weight calculator 120 calculates a weight using a random walk probability.
In detail, the method of calculating the weight using the random walk probability by the random walk weight calculator 120 will be described in comparison with the method of calculating the weight by a general bilateral filter. In the case of a general two-sided filter, weights are calculated using distance and color information between pixels.
Referring to Equation 1, a weight calculation method of a general two-side filter will be described. D p is a depth value of the current pixel p. f (b pq) is an absolute value of the distance difference between the pixel at point p and the pixel at point q. g (∥Ip-Iq∥) is a Gaussian color distribution between the pixels. k p is the normalization factor. In general, the two-sided filter using Equation 1 calculates the weights between the similar pixels and the distance between pixels is close, thereby preserving and clarifying the information of the boundary portion.
Unlike the weight calculation method of the general two-sided filter, the random walk weight calculation unit 120 calculates the weight using a random walk probability.
The block separator 110 may block the entire image into a predetermined size.
The image may be a color image including color pixels and may be a depth map image.
The weight may be calculated after dividing the image into block units. In the case of using the two-sided filter by dividing the data into block units and using the two-sided filter, the effect of noise may be reduced more than the case of segmentation by applying random walk probability to the entire image.
FIG. 2 is a diagram for explaining a center pixel and a boundary pixel in an entire area of an image when a weight is calculated in units of blocks according to the present invention.
Referring to FIG. 2 and
The weight using the random walk probability is obtained by subtracting the sum of the cost values from the boundary pixel to the central pixel in the initial probability 1. Therefore, the model P (i, j) representing the random walk probability may be expressed as in
In
3 is an exemplary diagram for explaining a cost value in an image boundary disambiguation apparatus according to the present invention.
Referring to FIG. 3 and
In FIG. 3, cost values between pixels i of the central block and
4 is an exemplary diagram illustrating a process of calculating a weight in an image boundary disambiguation apparatus according to the present invention.
In FIG. 4, it is assumed that a cost value of each block calculated according to
5 is an exemplary view for explaining a method of calculating a cost value when a path is 2 or more in the image boundary disambiguation apparatus according to the present invention.
6 is another exemplary view for explaining a method of calculating a cost value when a path is 2 or more in the image boundary disambiguation apparatus according to the present invention.
5 and 6, when calculating the sum of the cost values, there may be two or more paths from the boundary pixel to the center pixel. Each path may pass through another block where the cost values may be different so that the sum of the cost values from the same boundary pixel to the same center pixel may be different. If the sum of the cost values is different, the weights are different.
If there are two or more paths, the cost value is calculated by selecting one of the path having the maximum weight using the random walk probability or the path having the minimum cost value. For example, in FIG. 5, if the cost value is 0.1 for the route 1, the cost value is 0.2 for the
In FIG. 6, a gray circle is a color pixel, and a black circle is a border pixel. The middle white circle is the current pixel or the center pixel. The center may vary depending on which pixel is viewed as a reference, and the sum of the cost values may be calculated by setting the current pixel as the reference. Calculation of the cost value uses the RGB color difference of adjacent pixels, and FIG. 6 shows that the weight is different because the color pixels passing through the selected path are different when moving from the specific boundary pixel to the center pixel.
In detail, a path having a maximum random walk probability or a path having a minimum cost value may be determined by applying a Dirichlet Problem method. The Dirichlet plamb method is described with reference to Equations 5-8.
The Dirichlet plamb method is a method using a Laplace matrix. D [x] represents D n (i) in matrix form. In Equation 5, w ij represents a cost value between pixels, and Σw ij (x i -x j ) 2 can be converted into a matrix operation x T Lx. L is a Laplace matrix that represents the sum of the cost of each path and the cost of each path.
Specifically, the matrix L will be described with reference to Equation 6, i and j each pixel. d i is the sum of the respective path cost values that pixel i has with adjacent pixels. If i and j are equal, then matrix L is di.
If two pixels are close to each other, for example, if | ij | = 1, the matrix L becomes -w ij .
The matrix operations of Equation 5 may be expressed as Equation 7 by representing the elements. L border , x border , or border , means that the row's index in the matrix is the border pixel components, and L non - border , x non - border , or non-border, means the row's index in the matrix. Is the case that is not a boundary pixel component. B is the portion of the entire Laplace matrix that does not belong to L border and L non-border .
The derivative of Equation 7 with respect to X non - border can be derived. Equation 7 is an equation for finding a path in which the sum of cost values becomes the minimum in Equation 6.
As a minimum path solution, a method other than the Dirichlet plamb method may be used. For example, it is also possible to use the Graph-Cut method rather than the Dirichlet plab method, which is a method using Graph theroy.
The filtering unit 130 averages and filters the values of pixels of the adjacent image by using the weights calculated by the random walk weight calculator 120.
That is, the image boundary disambiguation apparatus 100 according to the present invention defines a minimum path for each pixel, defines a probability, and performs a filter using the defined probability.
The image boundary disambiguation apparatus 100 according to the present invention may be applied to an image and may also be applied to a depth map for a 3D image. When the present invention is applied to a depth map, the characteristics of the depth map can be improved.
Specifically, when the image boundary disambiguation device 100 according to the present invention is applied to a color image, a flattened color image in which the boundary is preserved may be obtained.
When the image boundary clearing device 100 according to the present invention is applied to a depth map image, the boundary mismatch problem can be solved and the accuracy of the depth value can be improved.
In addition, the image boundary disambiguation apparatus 100 according to the present invention reduces the sensitivity to noise by calculating a weight using a random walk in units of blocks.
In addition, the image boundary disambiguation apparatus 100 according to the present invention may make the boundary clear even when adjacent pixels have an ambiguous color, thereby obtaining a flattened image in which the boundary is preserved.
7 is a flowchart of an image boundary method according to a preferred embodiment of the present invention.
Referring to FIG. 7, the image boundary method is described. The block divider 110 divides an image into units of a predetermined size block (step S710). The constant size may vary depending on the setting, and may not necessarily be divided into constant sizes, but may be divided into sizes having periodic or aperiodic properties.
The random walk weight calculator 120 calculates a cost value (step S720) and calculates a weight by subtracting the cost value calculated in step 1 (step S730).
The filtering unit 130 calculates and averages the adjacent pixels using the calculated weights (S740).
When the image boundary clearing method according to the present invention is applied to a color image, a flattened color image in which the boundary is preserved can be obtained.
When the image boundary disambiguation method according to the present invention is applied to the depth map image, the boundary mismatch problem can be solved and the accuracy of the depth value can be improved.
In addition, the image boundary disambiguation method according to the present invention reduces the sensitivity to noise by calculating a weight using a random walk on a block basis.
In addition, the image boundary disambiguation method according to the present invention may make the boundary clear even when adjacent pixels have an ambiguous color, thereby obtaining a flattened image in which the boundary is preserved.
8 is an exemplary diagram for describing an example in which an image boundary disambiguation method according to the present invention is applied to a depth map.
Comparing the initial depth map of FIG. 8 with the depth map filtered by the image boundary disambiguation method according to the present invention, it can be seen that the depth map filtered by the image boundary disambiguation method according to the present invention has a clearer boundary.
Referring to Table 1, in the
9 is an exemplary diagram for explaining another example of applying the image boundary disambiguation method according to the present invention to a depth map.
9, there is a video image and a depth map is created based on the video image. Compared to the initial depth map, the depth map using the method proposed by the present invention can be seen that the boundary is clearer and the boundary information is preserved.
The image boundary disambiguation method according to the preferred embodiment of the present invention can be embodied as computer readable codes on a computer readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored.
Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like. Computer-readable code in a distributed fashion can be stored and executed. In addition, functional programs, codes, and code segments for implementing the present invention can be easily deduced by programmers of the art to which the present invention belongs.
It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.
Claims (16)
And a random walk weight calculation step of calculating a weight using a random walk probability.
And a filtering step of averaging and filtering the values of pixels of an adjacent image by using the weights.
The weight using the random walk probability is an initial probability of 1, which is a value obtained by subtracting a sum of cost values from a pixel at a boundary to a pixel at a central portion.
And the cost value is a value obtained by calculating a color difference using a Gaussian distribution between adjacent pixels.
Block dividing step of dividing the entire area of the image by a block unit of a predetermined size,
The weight calculation is an image boundary disambiguation method, characterized in that calculated in each of the divided block units.
The sum of the cost values is calculated by selecting one of the path having the maximum weight or the path having the lowest cost value when there are two or more paths from the boundary pixel to the center pixel. Image boundary disambiguation method, characterized in that the value.
The method of claim 1, wherein the path having the largest random walk probability or the path having the smallest cost value is calculated by applying a Dirichlet Problem method.
And the filtering is performed on a depth map of the image.
And a random walk weight calculator configured to calculate a weight using a random walk probability.
And a filtering unit which averages and filters values of pixels of an adjacent image by using the weights.
And the weight using the random walk probability is a value obtained by subtracting a sum of cost values from the pixel at the boundary to the pixel at the center at an initial probability 1.
And the cost value is a value obtained by calculating a color difference using a Gaussian distribution between adjacent pixels.
Further comprising a block divider for dividing the entire area of the image by a block unit of a predetermined size,
And calculating the weights, respectively, in each of the divided block units.
The sum of the cost values is calculated by selecting one of the path having the maximum weight or the path having the lowest cost value when there are two or more paths from the boundary pixel to the center pixel. Image boundary disambiguation device, characterized in that the value.
And calculating a path having a maximum random walk probability or a path having a minimum cost value by calculating a Dirichlet Problem device.
And the filtering is performed on a depth map of the image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020120002051A KR20130081080A (en) | 2012-01-06 | 2012-01-06 | Apparatus and method for color image boundary clearness |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020120002051A KR20130081080A (en) | 2012-01-06 | 2012-01-06 | Apparatus and method for color image boundary clearness |
Publications (1)
Publication Number | Publication Date |
---|---|
KR20130081080A true KR20130081080A (en) | 2013-07-16 |
Family
ID=48992929
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
KR1020120002051A KR20130081080A (en) | 2012-01-06 | 2012-01-06 | Apparatus and method for color image boundary clearness |
Country Status (1)
Country | Link |
---|---|
KR (1) | KR20130081080A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113545071A (en) * | 2019-03-12 | 2021-10-22 | Kddi 株式会社 | Image decoding device, image decoding method, and program |
-
2012
- 2012-01-06 KR KR1020120002051A patent/KR20130081080A/en not_active Application Discontinuation
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113545071A (en) * | 2019-03-12 | 2021-10-22 | Kddi 株式会社 | Image decoding device, image decoding method, and program |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102867288B (en) | Depth image conversion apparatus and method | |
US9445075B2 (en) | Image processing apparatus and method to adjust disparity information of an image using a visual attention map of the image | |
US9525858B2 (en) | Depth or disparity map upscaling | |
RU2423018C2 (en) | Method and system to convert stereo content | |
EP2293586A1 (en) | Method and system to transform stereo content | |
KR102464523B1 (en) | Method and apparatus for processing image property maps | |
US20140340486A1 (en) | Image processing system, image processing method, and image processing program | |
TW201432622A (en) | Generation of a depth map for an image | |
CN104994375A (en) | Three-dimensional image quality objective evaluation method based on three-dimensional visual saliency | |
EP1958149A2 (en) | Stereoscopic image display method and apparatus, method for generating 3d image data from a 2d image data input and an apparatus for generating 3d image data from a 2d image data input | |
KR20100135032A (en) | Conversion device for two dimensional image to three dimensional image and method thereof | |
EP2643810A1 (en) | Depth estimation based on global motion | |
TW201029443A (en) | Method and device for generating a depth map | |
KR101364860B1 (en) | Method for transforming stereoscopic images for improvement of stereoscopic images and medium recording the same | |
Ko et al. | 2D to 3D stereoscopic conversion: depth-map estimation in a 2D single-view image | |
US9519956B2 (en) | Processing stereo images | |
CN106530336B (en) | Stereo matching method based on color information and graph cut theory | |
Jung | A modified model of the just noticeable depth difference and its application to depth sensation enhancement | |
KR101797035B1 (en) | Method for converting overlaying area into 3D image and apparatus thereof | |
CN101710426B (en) | Method for tracking depth image | |
US9787980B2 (en) | Auxiliary information map upsampling | |
CN106683072B (en) | 3D image comfort level quality evaluation method and system based on PUP image | |
KR20130081080A (en) | Apparatus and method for color image boundary clearness | |
KR101760463B1 (en) | Method and Apparatus for correcting a depth map | |
CN102595167B (en) | Depth uniformization method and device for 2D/3D video conversion |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WITN | Withdrawal due to no request for examination |