WO2013105720A1

WO2013105720A1 - Device and method for analyzing quality of three-dimensional stereoscopic image

Info

Publication number: WO2013105720A1
Application number: PCT/KR2012/009003
Authority: WO
Inventors: 김아란; 김하윤; 정원석; 전병기; 김종옥; 최지훈; 박영환; 윤민주
Original assignee: 에스케이플래닛 주식회사; (주)리얼디스퀘어; 고려대학교 산학협력단
Priority date: 2012-01-10
Filing date: 2012-10-30
Publication date: 2013-07-18
Also published as: KR101393621B1; KR20130081835A

Abstract

The present invention relates to a device and method for analyzing the quality of a three-dimensional stereoscopic image, and the device comprises: a three-dimensional effect analysis unit for extracting depth values for each pixel per frame from depth information on a two-dimensional image, generating a depth information histogram indicative of a distribution chart of a depth map in accordance with a depth gray value for the extracted depth values per frame, and generating a three-dimensional value (DF) by analyzing the similarity between the depth information histogram and a uniform distribution histogram by using a Bhattacharyya coefficient; a stability analysis unit for generating parallax information by using the depth information on the two-dimensional image, obtaining seven coefficient values by using the depth information and the parallax information, and generating a final stability coefficient value (CF) through a linear combination; an error analysis unit for detecting an error that processes the background as an object or an error that processes the object as the background, by using the depth information on the two-dimensional image or other information excluding the depth information; and an analysis result unit for calculating a total evaluation score by giving a weight and a constant to each coefficient value on a three-dimensional effect, stability, and errors for each sequence image when the two-dimensional image is converted into a three-dimensional stereoscopic image, and providing a list according to the ranking with the total evaluation score for each sequence image.

Description

Apparatus and method for quality analysis of 3D stereoscopic images

The present invention relates to an apparatus and method for analyzing a quality of a 3D stereoscopic image. More particularly, how good a 3D stereoscopic image is by using depth map generated when the 2D image is converted into a 3D stereoscopic image. Display the degree of three-dimensional sense of whether you have a numerical value, calculate the seven factors that affect the stability of three-dimensional stereoscopic images, and indicate how severe visual fatigue is, and process the object in the background It is to be able to perform an objective quality evaluation on whether there is no error processing an as an object.

Recently, with increasing interest in 3D stereoscopic images, various types of stereoscopic image acquisition devices and display devices have been developed. One method for acquiring a stereoscopic image signal to display a stereoscopic image is to use a pair of left and right cameras when acquiring an image. This method has the advantage of displaying a natural three-dimensional image, but not only two cameras are required for image acquisition, but also problems caused by filming or encoding the acquired left and right images, and frame rates of the left and right images. There are disadvantages to solve the problems such as difference of the difference.

Another method for acquiring a 3D image signal is to convert a 2D image signal acquired by one camera into a 3D image signal. According to this method, since a three-dimensional image, i.e., a left image and a right image, is generated by performing predetermined signal processing on the acquired two-dimensional image (the original image), when acquiring a stereoscopic image signal using a left and right camera. The above-mentioned problem does not occur. However, since this method generates two images using one image, it is difficult to display a three-dimensional image having a natural and stable three-dimensional effect. Therefore, in converting a 2D video signal into a 3D video signal, it is very important to display a 3D video having a more natural and stable 3D effect by using the converted 3D video signal.

By the way, the dizziness, visual fatigue, etc. that viewers feel when they come in contact with 3D stereoscopic images are studied. Tools and devices for measuring its own quality for 3D stereoscopic images have not been developed yet.

The present invention has been made to solve the above-described problem, by using the depth map generated when converting a 2D image into a 3D stereoscopic image, the degree of the stereoscopic sense of how good the stereoscopic sense is Display the numerical value, obtain seven factor values that affect the stability of the 3D stereoscopic image, and indicate how severe the visual fatigue is, and whether there is no error processing the object as a background or processing the background as an object. There is a technical problem to provide an apparatus and method for quality analysis of three-dimensional stereoscopic images, which enables to perform an objective quality evaluation for.

According to an aspect of the present invention, there is provided an apparatus for quality analysis of a 3D stereoscopic image, comprising: an input unit configured to input an original sequence image and a depth map sequence image of a 2D image; Extract depth values for each pixel per frame from the depth information of the 2D image, and generate a depth information histogram indicating a depth map distribution map according to depth gray values for the extracted depth values for each frame. A three-dimensional analysis unit configured to generate a three-dimensional (DF) value by analyzing the similarity between the depth information histogram and the equal distribution histogram using a batcharya coefficient; Create parallax information using depth information on the 2D image, and use the depth information and parallax information between the range of parallax, the maximum value of parallax absolute value, the average value of parallax and the screen Seven coefficients for Average, Spatial Complexity, Depth Position of Screen and Distance Square, Temporal Complexity, and Mean Movement Square of Parallax Difference A stability analyzer for generating a final stability coefficient (CF) value by obtaining a linear combination; An error analysis unit that detects an error of processing a background as an object or an error of processing an object as a background by using depth information of the 2D image or using information other than the depth information; When converting the 2D image into a 3D stereoscopic image, the total evaluation score is calculated by giving weights and constants to the values of the stereoscopic sense, stability, and error of the sequence images for each sequence image. An analysis result unit having a total evaluation score and providing the list according to a ranking; And output coefficient values related to stereoscopic effect, stability, and error for each sequence image on the screen in numerical values and on a graph, and output a list according to the total evaluation score for each sequence image and the rank of each total evaluation score on the screen. It includes an output unit.

The analysis result unit may output a total evaluation score (QM) by calculating a linear combination of three-dimensional effect and stability with respect to the total evaluation score.

[Equation 1]

Here, QM represents the total total evaluation score of the 3D image, DF represents the stereoscopic value of the 3D image, CF represents the stability coefficient value of the 3D image, represents the weight, and represents the cost. Indicates.

The analysis result unit may determine the stereoscopic value DF as a logarithm of a histogram stereoscopic value and a parallax product according to Equation 2 below.

[Equation 2]

The three-dimensional analysis unit may further include: a histogram generator configured to generate a depth information histogram indicating a depth map distribution according to a depth gray value with respect to the depth information per frame; And a stereoscopic value calculation unit configured to calculate a stereoscopic value (DF) using the Battacharyya coefficient for the similarity between the depth map histogram and the uniform distribution histogram.

The histogram generator may convert depth values of each pixel per frame into a depth gray value of 0 to 255, and generate a depth information histogram indicating a distribution of depth gray values for each pixel in one frame. Can be generated.

In addition, the stereoscopic value calculation unit normalizes the frequency of the depth map histogram by dividing the number of pixels by the total number of pixels, and normalizes the i th value of the depth information histogram (x [i]). According to Equation 4, the i th value of the depth information histogram may be calculated by dividing the entire pixel value.

[Equation 4]

The stability analysis unit may extract depth values for each pixel per frame from a depth map of the 2D image and store depth information for each frame by dividing the depth values for each pixel. part; A parallax information generation unit generating parallax information using the generated depth map; And a stability calculator configured to obtain seven coefficient values using the depth information and parallax information, and to generate a final stability coefficient (CF) value by linear combination.

In addition, the stability calculator may calculate a final stability coefficient value CF for one frame by linear combination of the seven coefficient values according to Equation 12 below.

[Equation 12]

Here, ω1 to ω7 represent weight values for seven coefficient values.

In addition, the stability calculation unit calculates seven coefficient values according to Equation 10 for each frame for every frame in the same process of calculating the stability coefficient value CF for one frame, for each frame for every frame. The total stability factor (TCF) is arranged by arranging the seven coefficient values of the frame according to the following equation (13) (Total Compatibility Factor; TCF), and the weights of the seven coefficient values for each frame according to the following equation (14) It is possible to calculate the weighting factor (X) values arranged from (ω) values (ω1 to ω7).

[Equation 13]

[Equation 14]

Here, X represents a weighting coefficient value consisting of a 7x1 matrix, S represents a matrix vector for Nx1 subjective evaluation, and N represents the number of frames.

In addition, the error analysis unit, depth information for extracting the depth value for each pixel per frame from the depth map (depth map) for the two-dimensional image, the depth information for dividing and storing the depth value for each extracted pixel for each frame Storage unit; A parallax information generation unit generating parallax information using the generated depth map; And generating error coefficient values for depth reversal of processing an object as a background or processing a background as an object by using a depth value for each frame for the 2D image or using information other than the depth information. It includes an error calculation unit.

The error calculating unit may separate the background information and the object part using the depth value for each depth information per frame, and use the information other than the depth in the separated background part to determine the background object. Depth inversion to be processed can be detected to calculate the error coefficient value.

In addition, the error image calculating unit may be regarded as a background when a difference value of the difference image is smaller than a threshold value by using a difference image between a previous frame and a current frame of the 2D image as information other than the depth, and as an object if larger than the threshold value. The error count can be calculated.

The error calculator may detect an error using a motion vector with information other than the depth. The error calculator may set a global motion vector in a block unit and move in a block unit. When the calculated motion vector value differs from the motion vector of the background by a predetermined amount or more, the error coefficient value may be calculated by considering the object as an object.

On the other hand, the quality analysis method of the three-dimensional stereoscopic image of the present invention for achieving the above object, (a) inputting the original sequence image and depth map (depth map) sequence image for the two-dimensional image; (b) Depth information extracting depth values for each pixel per frame from the depth information of the two-dimensional image, and depth information indicating the depth map distribution according to the depth gray value for each extracted depth value for each frame Generating a histogram and analyzing the similarity between the depth information histogram and the uniform distribution histogram using a batcharya coefficient to generate a stereoscopic (DF) value; (c) generating parallax information using depth information of the 2D image, and using the depth information and parallax information, the range of parallax, the maximum value of the parallax absolute value, and the average value of parallax; Of the distance between the screen and the screen, the variance of the parallax, the variance of the screen and the distance square, the temporal complexity, and the average of the squares of the parallax differences. Obtaining seven coefficient values to produce a final stability coefficient (CF) value in a linear combination; (d) detecting an error of processing a background as an object or an error of processing an object as a background using depth information of the 2D image or information other than the depth information; And (e) calculating a total evaluation score by giving weights and constants to the three-dimensional, stability, and error coefficient values for each sequence image of the 2D image, or calculating the total evaluation score for each sequence image. Providing a list according to the ranking (Ranking).

On the other hand, the program for executing the quality analysis method of the three-dimensional stereoscopic image according to an embodiment of the present invention can be recorded in a computer-readable medium such as a CD or USB media.

As described above, the apparatus and method for quality analysis of 3D stereoscopic images of the present invention can objectively evaluate the quality of 3D stereoscopic images.

In other words, the quality of the depth level (depth) of the three-dimensional image, which is one of the factors for measuring the quality of the 3D stereoscopic image, can be confirmed numerically, and the seven factors for the stability (comfort), which is the factor for measuring the quality, are measured. Values can be calculated and evaluated, and the quality of 3D stereoscopic images can be confirmed by objectively evaluating error, which is a factor for measuring quality.

1 is a block diagram schematically showing a functional block of an apparatus for analyzing quality of a 3D stereoscopic image according to an exemplary embodiment of the present invention.

2 is a flowchart illustrating an overall operation of a method for analyzing quality of a 3D stereoscopic image according to an exemplary embodiment of the present invention.

3 is a control block diagram of a stereoscopic analysis unit according to an embodiment of the present invention.

4 is a flowchart illustrating a three-dimensional analysis method of the three-dimensional analysis unit according to an embodiment of the present invention.

5 is a diagram illustrating examples of depth values for each pixel for each frame generated from a 2D image according to an exemplary embodiment of the present invention.

FIG. 6 is a diagram illustrating an example of a histogram showing a distribution chart occupied by a depth value of a pixel per frame according to an exemplary embodiment of the present invention.

7 is a diagram illustrating an example in which the similarity between the depth information histogram and the uniform distribution histogram is shown in three-dimensional values according to an embodiment of the present invention.

FIG. 8 is a diagram illustrating examples of good and bad stereoscopic feelings for one frame of a 3D stereoscopic image based on the stereoscopic values calculated according to an embodiment of the present invention.

9 schematically illustrates a functional block of a stability analyzer according to an embodiment of the present invention.

10 is a flowchart illustrating a method for analyzing stability of a stability analyzer according to an exemplary embodiment of the present invention.

FIG. 11 is a diagram for describing a range of parallax and a maximum value of parallax according to an embodiment of the present invention.

12 is a diagram for describing a distance between an average value of parallax and a screen according to an exemplary embodiment of the present invention.

FIG. 13 is a diagram illustrating a spatial complexity of parallax according to an embodiment of the present invention. FIG.

14 is a view for explaining an example of calculating a coefficient value regarding a depth position according to an embodiment of the present invention.

15 schematically illustrates a functional block of an error analysis unit according to an embodiment of the present invention.

16 is a flowchart illustrating an error analysis method of an error analysis unit according to an exemplary embodiment of the present invention.

17 is a diagram illustrating an example of separating a background into an object using a depth value with respect to a 2D image according to an exemplary embodiment of the present invention.

FIG. 18 is a diagram illustrating an example of dividing an object portion and a background portion by using depth values for a manually converted image and an automatically converted image.

19 illustrates an example of using a difference image between a previous frame and a current frame of a 2D image according to an embodiment of the present invention.

20 is a diagram illustrating an example of detecting an error by considering a depth inversion as a value of a difference image is greater than a threshold value in a background part according to an exemplary embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Referring to FIG. 1, the apparatus 100 for analyzing quality of a 3D stereoscopic image according to an exemplary embodiment of the present invention may include an input unit 110, a stereoscopic analysis unit 120, a stability analysis unit 130, and an error analysis unit ( 140, an analysis result unit 150, an output unit 160, and the like.

The input unit 110 inputs an original sequence image and a depth information sequence image for a 2D image. At this time, when a sequential number such as "_001, _002, _003, ..." is attached to the end of the file name of the sequence image in one folder, all the images in the folder are loaded and input.

The stereoscopic analysis unit 120 extracts depth values for each pixel of each frame as shown in FIG. 5 from a depth map of the 2D image, and extracts depth gray for each extracted depth information for each frame. Depth histogram showing depth map distribution according to (depth gray) values, and the similarity between depth map histogram and uniform distribution histogram is calculated by Bhattacharyya Coefficient. ) To generate a QM factor. At this time, it can be seen that the stereoscopic value is lower than 0.3, for example, intermediate between 0.3 and 0.8, and higher than three-dimensional has a high three-dimensional effect.

The stability analyzer 130 generates parallax information using a depth map of the 2D image, and a range of parallax that affects stability of the 3D stereoscopic image using depth information and parallax information. (Range), the maximum value of the parallax absolute value (Max), the distance between the mean value of the parallax and the screen (Average), the spatial complexity of the parallax, the variance of the screen and the distance squared (Depth Position), the variance of the parallax difference Seven factors, such as Temporal Complexity and the mean of the square of the parallax difference, are obtained, and the final stability coefficient (CF) is generated by linear combination.

The error analysis unit 140 separates the depth information for each frame into a background part and an object part using the depth value, and processes the object in the background based on the threshold value. Detect an error in depth map reverse that treats the background as an object.

The analysis result unit 150 gives a weight and a constant for each coefficient value such as stereoscopic feeling, stability, and error for each sequence image when converting a 2D image into a 3D stereoscopic image, thereby giving 1 (bad) The total evaluation score is calculated and displayed from 5 to (good), and the total evaluation score for each sequence image is provided as a list according to the ranking. That is, the analysis result unit 150 outputs the total evaluation score QM by calculating the linear combination of the three-dimensional effect and the stability according to the following equation 1 for the final quality result value.

Equation 1

Here, QM represents a total total evaluation score of the 3D image, DF represents a stereoscopic value of the 3D image, and CF represents a stability coefficient value of the 3D image. In addition, denotes a weight (coefficient) represents a coefficient value for the stereoscopic value of the 3D image and a coefficient value for the stability of the 3D image can be obtained in the same manner as in Equations 12 to 14, and is a cost The final result indicates a value determined to adjust the value by adding a certain value so that the total evaluation score is between 0 and 5 points. In this case, in the case of the stereoscopic value, when the parallax increases, the point where the stereoscopic feeling no longer improves is generated, and the analysis result unit 150 analyzes the histogram stereoscopic value and the parallax according to the following equation (2) for the stereoscopic value DF. parallax) is determined by the logarithm of the product.

Equation 2

The output unit 160 outputs coefficient values related to stereoscopic feeling, stability, and error for each sequence image on the screen as numerical values and graphs, and lists the total evaluation scores for each sequence image and the rank of each total evaluation score. And so on. At this time, the output unit 160 displays each score for the sequence image in a polar graph so that the user can identify the score for each frame at a glance.

Referring to FIG. 2, the apparatus 100 for analyzing quality of a 3D stereoscopic image first receives a raw sequence image and a depth map sequence image through the input unit 110 (S210). ).

Accordingly, the 3D analysis unit 120 extracts depth values for each pixel of each frame as shown in FIG. 5 from a depth map of the 2D image, and extracts depth information for each extracted frame. Create a depth information histogram that shows the depth map distribution according to the depth gray value, and calculate the similarity between the depth map histogram and the uniform distribution histogram. Analyze using Bhattacharyya Coefficient to generate a three-dimensional value (QM Factor) (S220).

In addition, the stability analysis unit 130 generates parallax information using a depth map of the 2D image, and the parallax affecting the stability of the 3D stereoscopic image using the depth information and the parallax information. Range, the maximum value of the parallax absolute value (Max), the average value of the parallax and the distance between the screen (Average), the spatial complexity of the parallax, the variance of the screen and the distance squared (Depth Position), the parallax difference Seven factors, such as Temporal Complexity and the mean of the square of the parallax difference, are obtained, and the final stability coefficient (CF) is generated by linear combination (S230).

That is, when the user clicks the button for stability, the safety analysis unit 130 calculates and outputs a WeightSum of the average result value of the frames for the factors in the stability, and the result value for each frame is a text file. Save it so that you can see the result for each frame.

In addition, the error analysis unit 140 separates the depth information for each frame into a background part and an object part using the depth value, and sets the background of the object based on the threshold value. In operation S240, an error value for depth map reverse, which is processed as or is processed as an object, is generated.

That is, when the depth reversal is detected, the error analysis unit 140 outputs a depth map inverted as a thumbnail image, and if there is an error, the user selects a corresponding frame through the UI (User Interface) Only the result value can be excluded, and the inverted part shows how many areas are inverted and where they are inverted by white blocks of 8 * 8 pixels. In this case, the ratio of the white blocks to the total number of blocks in one frame is scored to generate an error value, and when there is more than 15 blocks of inversion in one frame, the abnormality of the corresponding frame image is displayed, and each frame has a threshold value or more. The frame in which the reverse phenomenon occurs can be displayed as an error frame.

Subsequently, the analysis result unit 150 calculates a total evaluation score from 1 (bad) to 5 (good) by giving weights and constants to respective values of stereoscopic sense, stability, and error for each sequence image. The total evaluation score for each sequence image is provided as a list according to a ranking (S250).

That is, the analysis result unit 150 outputs the result values of the three-dimensional feeling, stability and error of each sequence image in numerical values and graphs, and outputs the original sequence image and the depth sequence image as thumbnails, and Print the resulting thumbnail.

Therefore, the user can check each value of the stereoscopic sense, stability and error of the original sequence image, and can check the total evaluation score on the overall quality.

[Detailed Structure and Operation of Stereoscopic Analysis]

As shown in FIG. 3, the stereoscopic analysis unit 120 of the present invention includes a histogram generator 310, a stereoscopic value calculator 320, and a stereoscopic value storage unit 330.

Here, the depth information for each frame of the 2D image is obtained by extracting depth values for each pixel of each frame as shown in FIG. 5 from the depth map of the 2D image. The depth map is a data structure that stores depth values of each pixel per frame for the 2D image. 5 is a diagram illustrating examples of depth values for each pixel for each frame generated from a 2D image according to an exemplary embodiment of the present invention. As shown in FIG. 5, the 3D analysis unit 120 extracts depth values for each pixel per frame from the depth map as shown in (A), and extracts depth values for each extracted pixel as shown in (B). Save by temporarily. At this time, the three-dimensional analysis unit 120 has a storage function for temporarily storing the generated depth information.

The histogram generator 310 generates a depth information histogram representing a depth map distribution according to the depth gray value with respect to the extracted depth information per frame. That is, the histogram generator 120 converts the depth values of each pixel per frame into depth gray values of 0 to 255 as the horizontal axis, as shown in FIG. 6, and sets each depth value in one frame. The histogram with the vertical axis showing the distribution of how much is distributed. FIG. 6 is a diagram illustrating an example of a histogram showing a distribution chart occupied by a depth value of a pixel per frame according to an exemplary embodiment of the present invention.

The stereoscopic value calculator 320 calculates the similarity between the depth map histogram and the uniform distribution histogram using a Battacharyya Coefficient, and outputs the similarity between the depth map histogram and the stereoscopic value DF. Here, the equal distribution histogram means an ideal depth information histogram in which the depth values of the depth map are variously distributed so that a three-dimensional appearance is high. Accordingly, the stereoscopic analysis unit 120 of the present invention stores an equal distribution histogram which is a reference for comparing with the depth information histogram of each pixel per frame.

That is, the stereoscopic value calculating unit 320 normalizes the number of pixels so that the sum of the depth information histogram and the uniform distribution histogram values is 1, and normalizes the depth information according to the following equation (3) using the Batcharya coefficient. The similarity between the information histogram and the uniform distribution histogram is calculated as a three-dimensional value (DFHistogram).

Equation 3

In Equation 3, is a value of the i-th index of the normalized depth information histogram, and is a value of the i-th index of the normalized uniformly distributed histogram.

Here, the Batcharya coefficient compares the previous frame with the current frame, finds the color similarity between the two frames, and recognizes and tracks the most similar pixels as the same pixels as the previous frame. The closer to 1, the higher the correlation between the two frames, and the closer to -1, the lower the correlation. That is, the closer the stereoscopic value is to 1, the more similar it is, and the closer to -1, the different.

First, the stereoscopic analysis unit 120 according to an embodiment of the present invention receives depth information for each frame of the 2D image through the input unit 110 (S410).

At this time, the 3D analysis unit 120 extracts depth values for each pixel per frame from the depth map of the 2D image, and extracts depth values for each extracted pixel for each frame. Save it temporarily.

Subsequently, the three-dimensional analysis unit 120 performs depth map distribution according to depth gray values through the histogram generator 310 with respect to depth information for each frame received through the input unit 110. A depth information histogram is generated (S420).

That is, the histogram generator 310 converts the depth values of each pixel per frame into depth gray values of 0 to 255 values as shown in FIG. 6, and sets the horizontal axis, and in each frame in one frame. It is to generate a depth information histogram with a vertical axis indicating a distribution of depth gray values.

Subsequently, the stereoscopic analysis unit 120 determines the similarity between the depth map histogram and the uniform distribution histogram through the stereoscopic value calculator 320 using the Bhattacharyya Coefficient. It calculates it as a value (QM Factor) (S430).

At this time, the three-dimensional value calculation unit 320 is normalized by the number of pixels so that the sum of the depth information histogram and the equal distribution histogram value is 1, respectively, and normalized as shown in Equation 1 by using the Batcharya coefficient. By multiplying each index value of the depth information histogram by each index value of the normalized uniform distribution histogram and adding the roots together, as shown in FIG. 7, the similarity of the depth information histogram and the uniform distribution histogram as a three-dimensional value (DFHistogram). To calculate. 7 is a diagram illustrating an example in which the similarity between the depth information histogram and the uniform distribution histogram is shown in three-dimensional values according to an embodiment of the present invention.

That is, the stereoscopic value calculator 320 normalizes the frequency of the depth map histogram by dividing the frequency by the number of pixels. In this case, the total pixel total may be 2,073,600 pixels, for example, in 1920 * 1080 Full HD. The normalization (x [i]) of the i-th value of the depth information histogram may be calculated by dividing the i-th value of the depth information histogram by all pixel values according to Equation 4 below.

Equation 4

Therefore, the sum from i th to 255 of the normalized depth histogram value is 1, and the first index of the normalized depth histogram for the depth gray value from 0 to 255. The rooted value is multiplied by the value of (index) (x1) and the value of the first index (y1) of the normalized uniform distribution histogram, and the value of the second index (x2) of the normalized depth information histogram (x2). ) Is multiplied by the value of the second index (y2) of the normalized uniform distribution histogram, and the root is obtained, and the value of the 256th index (x256) of the depth histogram normalized by the same process (x256) By multiplying the value of the 256th index (y256) of the distribution histogram, the rooted values are added together to calculate the stereoscopic value (DFHistogram).

The stereoscopic analysis unit 120 displays the stereoscopic value for each frame calculated by the stereoscopic value calculating unit 320 on the screen as shown in FIG. 8, or the stereoscopic value storage unit 330 for later use. Save to (S440).

That is, the stereoscopic value calculator 320 outputs the stereoscopic values for each frame, calculated as shown in Equation 1, on the screen together with the frame and the histogram, so that the user has a good stereoscopic effect and stereoscopic effect as shown in FIG. 8. This is a bad example. FIG. 8 is a diagram illustrating examples of good and bad stereoscopic feelings for one frame of a 3D stereoscopic image based on the stereoscopic values calculated according to an embodiment of the present invention.

Therefore, according to the stereoscopic analysis unit 120 according to an embodiment of the present invention, it can be confirmed by numerical value how good the 3D stereoscopic image has a stereoscopic feeling.

[Detailed Configuration and Operation of Stability Analysis Unit]

When the stability analysis unit 130 receives the original sequence image and the depth map sequence image through the input unit 110, the stability analysis unit 130 stores the input depth map sequence image in the depth information storage unit 910.

That is, the depth information storage unit 910 extracts the depth values for each pixel per frame from the depth map of the 2D image as shown in FIG. 5, and the depth for each extracted pixel. It has a function to store values separately for each frame as shown in (b).

Subsequently, the parallax information generator 920 generates parallax information using the generated depth map. A person perceives the three-dimensional or distance of an object through two eyes. When looking at an object, different images are generated through two eyes, and by appropriately synthesizing the images, a person perceives a three-dimensional sense or a sense of distance of the object. Accordingly, the 3D stereoscopic image generates two images in consideration of visual differences between left and right eyes from one 2D image in consideration of the cognitive characteristics of the person. This visual difference between the eyes is parallax, and creates a parallax using a depth map.

Subsequently, the stability calculator 930 uses depth information and parallax information to determine the range of parallax that affects the stability of the 3D stereoscopic image, the maximum value of the parallax absolute value, the average value of the parallax, and the screen. Seven factors: distance, spatial complexity, variance of screen and distance squares, temporal complexity, and mean movement squared ) Is then used to generate the final coefficient of stability (CF) by linear combination.

First, the stability analyzer 130 according to an embodiment of the present invention receives depth information for each frame of the 2D image through the input unit 110 (S1010).

In this case, the stability analysis unit 130 extracts depth values for each pixel per frame from the depth map of the 2D image, and extracts the depth values for each extracted pixel for each frame. Save separately.

Subsequently, the stability analyzer 130 generates parallax information through the parallax information generator 920 using the depth map for each frame (S1020).

That is, the parallax information generation unit 920 generates parallax information about visual differences of eyes used when generating two images considering visual differences between left and right eyes from one 2D image.

Subsequently, the stability analyzer 130 uses the depth information and the parallax information to determine the range of parallax, the maximum value of the parallax absolute value, the average value of the parallax, and the distance between the screens through the stability calculator 930. Average, Spatial Complexity, Depth Position of the Screen and Distance Squares, Temporal Complexity, and Seven Movements on the Mean Square of the Parallax Difference. The value is calculated (S1030).

That is, the stability calculator 930 calculates the coefficient value CFRange of the range of parallax according to Equation 5 as the first coefficient value.

Equation 5

Herein, the range of parallax is the value where the disparity is most forwarded based on the screen of the depth map for each frame, as shown in FIG. 11. Let max be the maximum value of the negative (-) value of the disparity value when the screen value is 0, and min (the largest value of the disparity value as the positive value). The larger the range value, the larger the parallax width, which may affect stability.

Also, disp (x, y) is a value of how much the x, y coordinate values of the original image are moved in the left image, and the gray value of the x, y coordinates of the depth map is 256 (the total depth of the image). subtracted from the gray level value (256-depthmap (x, y coordinate gray value) minus the gray value of the screen surface [(256-depthmap (x, y coordinate gray value))-screen gray value]) That is, information on how much difference between left and right images exists based on the screen surface Figure 11 is a range of parallax and a maximum value of parallax according to an embodiment of the present invention. It is a figure for explaining.

In addition, the stability measuring unit 150 calculates the coefficient value CFMax regarding the maximum value Max of the parallax absolute value according to Equation 6 as the second coefficient value.

Equation 6

Here, as shown in FIG. 11, the maximum value MAX is a value in which a disparity has the most forward value based on a screen in the depth map for each frame (disp (x, y), that is, negative (−). ) Value, so the absolute value is applied.) The higher the MAX value, the more images appear ahead of the screen, which can affect stability.

In addition, the stability calculator 930 calculates a coefficient value CFaverage of the distance between the average value of the parallax and the screen as shown in FIG. 12 as the third coefficient value.

Equation 7

Here, dispdisplay represents a disp. Value in a display and is zero because there is no difference in the screen.

12 is a diagram illustrating a distance between an average value of parallax and a screen according to an exemplary embodiment of the present invention. As shown in FIG. 12, the larger the Average value, the farther from the screen surface may affect stability. In addition, when the absolute value is subtracted, if the Average value becomes negative (-), the area that protrudes more than the screen surface has a bad effect on stability. Here, dispdisplay represents the disp. Value in the display and is zero because there is no difference in the screen.

In addition, the stability calculator 930 calculates a coefficient value CFSpatialComplexity as a fourth coefficient value with respect to spatial complexity as shown in FIG. 13 according to Equation 8 below.

Equation 8

Here, E (disp (x, y)) represents a disparity value in the average plane of (x, y) coordinates, and the dispersion of parallax is a dispersion value in the average plane as shown in FIG. As the density is more densely distributed in the average point, dizziness may be reduced than in the same image (the image having a large dispersion value in the average point). FIG. 13 is a diagram illustrating a spatial complexity of parallax according to an embodiment of the present invention. FIG.

In addition, the stability calculator 930 calculates the coefficient value CFDepthPosition regarding the depth position as shown in FIG. 14 as the fifth coefficient value.

Equation 9

Here, dispdisplay represents the disp. Value in the display and is zero because there is no difference in the screen.

That is, the coefficient value CFDepthPosition for the depth position is a variance value of the object distance squared from the screen as shown in FIG. 14. 14 is a view for explaining an example of calculating a coefficient value regarding a depth position according to an embodiment of the present invention.

In addition, the stability calculator 930 calculates a coefficient value (CFTemporalComplexity) regarding the variance of the parallax difference according to Equation 10 as the sixth coefficient value.

Equation 10

Where dispi (x, y) represents the disparity value of the (x, y) coordinate in the current i frame, and dispi-1 (x, y) represents (x, y in the previous i-1 frame Diff (x, y) represents the difference (diff.) Between the parallax value of the current i frame and the disparity value of the previous i-1 frame, and E (diff (x) , y)) represents the difference (diff.) value of parallaxes on the screen side.

In addition, the stability calculator 930 calculates the coefficient value CFSceneMovement regarding the mean (Scene Movement) of the square of the parallax difference according to Equation 11 as the seventh coefficient value.

Equation 11

Where dispi (x, y) represents the disparity value of the (x, y) coordinate in the current i frame, and dispi-1 (x, y) represents (x, y in the previous i-1 frame ) Shows the disparity value of the coordinate, and diff (x, y) indicates the difference between the disparity value of the current i frame and the disparity value of the previous i-1 frame.

Subsequently, the stability calculator 930 linearly combines the seven coefficient values according to Equation 12 to calculate a final stability factor (CF) for one frame (S1040).

Equation 12

Here, ω1 to ω7 represent weight values for seven coefficient values.

In this case, the stability analyzer 130 calculates seven coefficient values according to Equation 10 for each frame for all the frames in the same process of calculating the stability coefficient value CF for one frame, for all the frames. The total stability factor is obtained by arranging the seven coefficient values of each frame according to the following equation (13) (Total Compatibility Factor; TCF). Calculate a weighting factor X value (ω1 to ω7) arranged with weights ω values. That is, the stability calculator 930 determines the weight coefficient of the linear combination so that the correlation with the subjective evaluation is high by means of the Least Mean Square method.

Equation 13

Equation 14

Therefore, according to the stability analysis unit 130 according to an embodiment of the present invention, it can be confirmed by numerical value how good stability (comfort) of the 3D stereoscopic image.

[Detailed Configuration and Operation of Error Analysis Unit]

When the error analysis unit 140 receives the original sequence image and the depth map sequence image through the input unit 110, the error analysis unit 140 stores the input depth map sequence image in the depth information storage unit 1510.

That is, the depth information storage unit 1510 extracts depth values for each pixel per frame from the depth map of the 2D image as shown in FIG. 5, and the depth for each extracted pixel. The values are divided and stored for each frame as shown in (b).

Subsequently, the parallax information generation unit 1520 generates parallax information using the generated depth map. The 3D stereoscopic image generates two images in consideration of visual differences between left and right eyes from one 2D image in consideration of human cognitive characteristics. This visual difference between the eyes is parallax, and creates a parallax using a depth map.

The error calculator 1530 divides the generated depth information for each frame into a background part and an object part by using a depth value, and sets an object based on a threshold value. Error coefficients for depth inversion or background processing as objects.

In this case, the error calculator 1530 separates the background and the object by using the object having a depth value greater than that of the background in each horizontal line of depth information.

In addition, the error calculator 1530 sets the depth value of the background as a threshold value for each horizontal line in the depth information, and considers the object larger than the threshold value for each horizontal line as the object. To separate the objects from the background.

In addition, the error calculator 1530 detects a depth reversal for processing the object in the background as the depth value of the background among the separated object parts, and calculates an error count value.

In addition, the error calculator 1530 may detect a depth reversal of processing the background as an object using information other than the depth in the separated background to calculate an error coefficient value. Here, the information other than the depth is to use the difference image of the 2D image, and if the difference value of the difference image is smaller than the threshold value using the difference image of the previous frame and the current frame of the 2D image, it is regarded as the background, If it is large, it is regarded as an object. This assumes that the movement of the camera is not very dynamic. Since the background part is almost unchanged and only the object part is changed, the value of the difference image of the background part is small and the value of the difference image of the object part is relatively large. will be. Accordingly, the error calculator 1530 calculates an error count value by considering the depth inversion when a part of the background image having a difference value greater than the threshold exists in the background part.

The error calculator 1530 may detect an error using a motion vector with information other than depth. The error calculation unit 1530 sets a global motion vector in a block unit and sets a block unit in a block unit. When the calculated motion vector value is more than a predetermined distance from the motion vector of the background, the motion vector is regarded as an object and an error coefficient value is calculated.

Referring to FIG. 16, the error analysis unit 140 according to an embodiment of the present invention first receives depth information for each frame of a 2D image through the input unit 110 as shown in FIG. 5. (S1610).

That is, the error analysis unit 140 extracts depth values for each pixel per frame from the depth map of the 2D image, and extracts the depth values for each pixel from each frame as shown in FIG. 5. The data is stored in the depth information storage unit 1510.

Subsequently, the error analysis unit 140 uses the depth value through the error calculation unit 1530 for depth information for each frame, as shown in FIG. 17, as shown in FIG. 17. It is separated into the (object) part (S1620). 17 is a diagram illustrating an example of separating a background into an object using a depth value with respect to a 2D image according to an exemplary embodiment of the present invention. 18 illustrates an example of separating the object part and the background part by using depth values for the manually converted image and the automatically converted image.

FIG. 17A illustrates a frame of a left image for generating a 3D stereoscopic image, and FIG. 17B illustrates a depth map of the image of FIG.

In general, since the brightness of the pixels representing the object is brighter than the brightness of the pixels representing the background, as shown in FIG. 17B, the depth information of the pixels representing the object is greater than the depth information of the pixels representing the background. Able to know.

However, area A should be included in the background although the depth information is large, and area B should be included in the object although the depth information is small.

When the depth map is generated by manually converting the image of FIG. 17A, the depth map is illustrated in FIG. 18A, and the depth map is automatically generated when the depth map is automatically converted to FIG. 18B. If the object and the background are separated by using the depth information of each depth map, it becomes (a ') and (b') of FIG.

That is, when each pixel is automatically converted into a background or an object only by the depth information, the area A of FIG. 17 is incorrectly separated into an object because the depth information is large although the area A should be separated into the background, and the area B may be separated into an object. Because of its small size, it can be incorrectly separated into the background.

Therefore, when comparing the automatically converted image of FIG. 18 (b ') and the manually converted image of FIG. 18 (a'), depth inversion occurs in some areas (A and B of FIG. 17), thereby incorrectly separating an object and a background. It can be seen that.

That is, the error calculator 1530 separates the background and the object by using the object having a depth value greater than that of the background in each horizontal line of depth information of the 2D image.

In addition, the error calculator 1530 sets a threshold value for each horizontal line in the depth information, and considers the object larger than a threshold value among the depth values for the background and the object for each horizontal line. To separate the objects from the background.

Also, as shown in FIG. 19, the error calculating unit 1530 uses the difference image between the previous image (frame) and the current image (frame) as information other than the depth, and if the difference image value is smaller than the threshold value. If it is regarded as a background and if it is larger than the threshold, it can be regarded as an object, so that the background is treated as an object or an error in which the object is processed as a background can be detected. 19 illustrates an example of using a difference image between a previous frame and a current frame of a 2D image according to an embodiment of the present invention.

Specifically, the image (a) of FIG. 19 is a current frame image, the image (b) of FIG. 19 is a previous frame image, and the image (c) of FIG. 19 represents a difference image obtained by subtracting the previous frame image from the current frame image. .

The difference image (Fig. 19 (c) image) is a difference occurs when the outline portion of the object (person) compares the current frame and the previous frame in accordance with the movement of the object, the difference in depth information occurs in this portion It can be seen that it is displayed in color.

The error calculator 1530 extracts an area where depth reversal occurs by comparing a depth map of an object or a background separated by the object / background separation unit and a difference image generated from the difference image.

Subsequently, the error analysis unit 140 detects a depth reversal through which the object is processed as the background or the background is processed as the object through the error calculation unit 1530 (S1630).

That is, the error calculator 1530 may determine the depth in which the object is processed as the background in the image (a) in which the background and the object are separated as shown in FIG. 20, as the depth value of the background exists among the separated object parts. The reversal (inner circle) is detected and the error count is calculated.

In addition, the error analysis unit 140 may detect an error in which the background is processed as an object by using the difference image as information other than the depth in the separated background part. That is, the error analysis unit 140, if the camera movement is not very dynamic, since the background portion is almost unchanged and only the object portion has a change in the image, the value of the difference image (b) of the background portion is small and the object (object) As shown in FIG. 20, when the value of the portion is greater than the threshold value in the background portion (black portion) as shown in FIG. 20, the value of the portion is regarded as depth inversion, and an error is detected. will be. FIG. 20C is a diagram illustrating an example of detecting an error by considering the depth inversion B as the value of the difference image b is larger than the threshold value in the background part.

That is, as shown in FIG. 20, the error analysis unit 140 compares the (a) image, which is the object or background image separated into the object / background, and the (b) image, which is the difference image.

The error analysis unit 140 estimates the outline of the object from the difference image (image (b) of FIG. 20), and generates an area in which a pixel separated from the background included in the object is inverted in depth (FIG. c) B area of the image).

When the error analysis unit 140 detects a pixel separated into a background among the pixels included in the object as an area occurring in the depth reversal, the depth information of the difference image pixel corresponding to the pixel constituting the separated object is inverted depth. Pixels less than the threshold value are extracted to an area where depth inversion occurs, and pixels having depth information of a difference image pixel corresponding to pixels constituting the separated background greater than or equal to the depth inversion threshold are extracted to an area where depth inversion occurs.

In addition, the error analysis unit 140 of the 3D stereoscopic image may detect a region where depth inversion occurs using a motion vector.

The motion vector calculation divides the frame into block units, calculates a motion vector of the block, and the error analysis unit 140 uses the calculated motion vector to determine whether the motion vector of the block constituting the separated background is a reference motion vector. The larger block is extracted to the area where depth reversal occurs.

Therefore, an error in the case where each pixel is separated into a background or an object using only depth information may be detected using a difference image or a motion vector, and the region in which the error is detected may be re-separated.

In addition, the error analysis unit 140 detects an error using a motion vector with information other than depth. The error analysis unit 140 sets a global motion vector in a block unit and sets a block unit. When the calculated motion vector differs from the background motion vector by a certain degree, the motion vector is regarded as an object and an error is detected.

Subsequently, the error analysis unit 140 displays the error detection on the screen or stores the error detection unit in operation S1640.

Therefore, according to the error analysis unit 140 according to an embodiment of the present invention, the quality of the 3D stereoscopic image can be confirmed by objectively evaluating the quality of the error, which is one factor for measuring the quality of the 3D stereoscopic image. have.

According to an embodiment of the present invention, using the depth map generated when converting a two-dimensional image to a three-dimensional stereoscopic image, the degree of three-dimensionality as to how good the three-dimensional image has a three-dimensional image as a numerical value Display and determine how severe the visual fatigue is by obtaining seven factor values that affect the stability of 3D stereoscopic images, and whether there is no error processing the object as the background or the background as the object in the depth information. An apparatus and method for analyzing quality of a 3D stereoscopic image may be implemented to allow an objective quality evaluation to be performed.

As such, those skilled in the art will appreciate that the present invention can be implemented in other specific forms without changing the technical spirit or essential features thereof. Therefore, the above-described embodiments are to be understood as illustrative in all respects and not as restrictive. The scope of the present invention is shown by the following claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included in the scope of the present invention. do.

The present invention uses a depth map generated when converting a two-dimensional image to a three-dimensional stereoscopic image, and displays the degree of three-dimensionality as to how good the three-dimensional image has a three-dimensional image, or three-dimensionally. Seven factor values that affect the stability of stereoscopic images are obtained to indicate how severe the visual fatigue is, or to provide an objective quality assessment of whether there is no error in processing the background or object in the depth information. It can be used in the apparatus and method for analyzing quality of 3D stereoscopic images.

Claims

An input unit for inputting an original sequence image and a depth map sequence image of a 2D image;

Extract depth values for each pixel per frame from the depth information of the 2D image, and generate a depth information histogram indicating a depth map distribution map according to depth gray values for the extracted depth values for each frame. A three-dimensional analysis unit configured to generate a three-dimensional (DF) value by analyzing the similarity between the depth information histogram and the equal distribution histogram using a batcharya coefficient;

Create parallax information using depth information on the 2D image, and use the depth information and parallax information between the range of parallax, the maximum value of parallax absolute value, the average value of parallax and the screen Seven coefficients for Average, Spatial Complexity, Depth Position of Screen and Distance Square, Temporal Complexity, and Mean Movement Square of Parallax Difference A stability analyzer for generating a final stability coefficient (CF) value by obtaining a linear combination;

An error analysis unit that detects an error of processing a background as an object or an error of processing an object as a background by using depth information of the 2D image or using information other than the depth information;

When converting the 2D image into a 3D stereoscopic image, the total evaluation score is calculated by giving weights and constants to the values of the stereoscopic sense, stability, and error of the sequence images for each sequence image. An analysis result unit having a total evaluation score and providing the list according to a ranking; And

Coefficient values for stereoscopic feeling, stability, and error are output on the screen as numerical values and graphs for each sequence image, and a list according to the total evaluation score and the ranking of each total evaluation score for each sequence image is output on the screen. An output unit;

3D stereoscopic image quality analysis apparatus comprising a.
The method of claim 1,

The analysis result unit is a three-dimensional, characterized in that for outputting the total evaluation score (QM) by calculating a linear combination of the three-dimensional sense and stability with respect to the total evaluation score according to the following equation (1) 3D image quality analysis device.

[Equation 1]

Here, QM represents the total total evaluation score of the 3D image, DF represents the stereoscopic value of the 3D image, CF represents the stability coefficient value of the 3D image, represents the weight, and represents the cost. Indicates.
The method of claim 2,

And the analysis result unit determines the stereoscopic value DF as a logarithm of a histogram stereoscopic value and a parallax product according to Equation 2 below.

[Equation 2]
The method of claim 1,

The stereoscopic analysis unit may include: a histogram generator configured to generate a depth information histogram indicating a depth map distribution according to a depth gray value with respect to the depth information per frame; And a stereoscopic value calculation unit configured to calculate a stereoscopic value (DF) by using a batatacharyya coefficient for the similarity between the depth map histogram and the uniform distribution histogram. 3D stereoscopic image quality analysis device.
The method of claim 4, wherein

The histogram generator converts depth values of each pixel per frame into depth gray values of 0 to 255, and generates a depth information histogram representing a distribution of depth gray values for each pixel in one frame. 3D stereoscopic image quality analysis device, characterized in that.
The method of claim 4, wherein

The stereoscopic value calculation unit normalizes the frequency of the depth map histogram by dividing the frequency by the number of pixels, and performs the following mathematical expression on the normalization (x [i]) of the i th value of the depth information histogram. And an i-th value of the depth information histogram is calculated by dividing the entire pixel value according to Equation 4.

[Equation 4]
The method of claim 1,

The stability analysis unit may include: a depth information storage unit for extracting depth values for each pixel per frame from a depth map of the 2D image, and storing the extracted depth values for each pixel for each frame; A parallax information generation unit generating parallax information using the generated depth map; And a stability calculator configured to obtain seven coefficient values using the depth information and parallax information, and generate a final stability coefficient (CF) value by linear combination.

3D stereoscopic image quality analysis apparatus comprising a.
The method of claim 7, wherein

The stability calculator calculates a final stability coefficient (CF) for one frame by linear combination of the seven coefficient values according to Equation 12 below. Device.

[Equation 12]

Here, ω1 to ω7 represent weight values for seven coefficient values.
The method of claim 8,

The stability calculation unit calculates seven coefficient values according to Equation 10 for each frame for all the frames in the same process of calculating the stability coefficient value CF for one frame. The total stability factor (TCF) is obtained by arranging the branch coefficient values as the number of frames according to the following equation (13), and the weight (ω) for the seven coefficient values for each frame according to the following equation (14). A quality analysis apparatus for a three-dimensional stereoscopic image, characterized by calculating a weighting coefficient (X) values arranged from the values (ω1 to ω7).

[Equation 13]

[Equation 14]

Here, X represents a weighting coefficient value consisting of a 7x1 matrix, S represents a matrix vector for Nx1 subjective evaluation, and N represents the number of frames.
The method of claim 1,

The error analysis unit extracts depth values for each pixel per frame from a depth map of the 2D image, and stores the depth information for each frame by dividing the depth values for each pixel. ; A parallax information generation unit generating parallax information using the generated depth map; And generating error coefficient values for depth reversal of processing an object as a background or processing a background as an object by using a depth value for each frame for the 2D image or using information other than the depth information. An error calculation unit;

3D stereoscopic image quality analysis apparatus comprising a.
The method of claim 10,

The error calculating unit divides the depth information per frame into a background part and an object part using a depth value, and processes the background as an object using information other than depth in the separated background part. Detecting depth reversal to calculate the error coefficient value of the three-dimensional stereoscopic image analysis device.
The method of claim 10,

The error image calculation unit uses the difference image between the previous frame and the current frame of the 2D image as information other than the depth to regard the background as a value when the difference image is smaller than a threshold value, and to regard the object as an object when the difference value is larger than the threshold value. 3D stereoscopic image quality analysis device, characterized in that for calculating an error coefficient value.
The method of claim 10,

The error calculator may detect an error using a motion vector using information other than the depth. The error calculation unit may set a global motion vector in a block unit, and set the motion vector in a block unit. And calculating an error coefficient value when the calculated motion vector value differs from the background motion vector by a predetermined amount, and calculates an error coefficient value.
(a) inputting an original sequence image and a depth map sequence image for the 2D image;

(b) Depth information extracting depth values for each pixel per frame from the depth information of the two-dimensional image, and depth information indicating the depth map distribution according to the depth gray value for each extracted depth value for each frame Generating a histogram and analyzing the similarity between the depth information histogram and the uniform distribution histogram using a batcharya coefficient to generate a stereoscopic (DF) value;

(c) generating parallax information using depth information of the 2D image, and using the depth information and parallax information, the range of parallax, the maximum value of the parallax absolute value, and the average value of parallax; Of the distance between the screen and the screen, the variance of the parallax, the variance of the screen and the distance square, the temporal complexity, and the average of the squares of the parallax differences. Obtaining seven coefficient values to produce a final stability coefficient (CF) value in a linear combination;

(d) detecting an error of processing a background as an object or an error of processing an object as a background using depth information of the 2D image or information other than the depth information; And

(e) calculate the total evaluation score by giving weights and constants to the values of the three-dimensional, stability, and error coefficients for each sequence image of the 2D image, or rank the total evaluation score for each sequence image Providing as a list according to (Ranking);

Quality analysis method of the three-dimensional stereoscopic image comprising a.
The method of claim 14,

In the step (e), the total evaluation score is calculated as a linear combination of three-dimensional feeling and stability according to the following Equation 1, characterized in that the output of the total evaluation score (QM), characterized in that 3D stereoscopic image quality analysis method.

[Equation 1]

Here, QM represents the total total evaluation score of the 3D image, DF represents the stereoscopic value of the 3D image, CF represents the stability coefficient value of the 3D image, represents the weight, and represents the cost. Indicates.
The method of claim 15,

In the step (e), the stereoscopic value DF is determined by a logarithm function of a histogram stereoscopic value and a parallax (parallax) product according to Equation 2 below.

[Equation 2]
The method of claim 14,

The step (b) may include: (a) generating a depth information histogram indicating a depth map distribution according to depth gray values for the depth information per frame; And (b) calculating a similarity between the depth map histogram and the uniform distribution histogram by using a battacharyya coefficient. Quality analysis method of 3D stereoscopic images.
The method of claim 17,

In step (a), the depth information of each pixel per frame is converted into a depth gray value of 0 to 255 values, and a depth information histogram representing a distribution of depth gray values for each pixel in one frame is obtained. Quality analysis method of the three-dimensional stereoscopic image, characterized in that for generating.
The method of claim 17,

In the step (b), the frequency of the depth map histogram is normalized by dividing the frequency of the depth map histogram by the number of pixels, and for the normalization (x [i]) of the i th value of the depth information histogram, And analyzing the i th value of the depth information histogram by the total pixel value according to Equation 4 to calculate the quality of the 3D stereoscopic image.

[Equation 4]
The method of claim 14,

In the step (c), the depth values of each pixel per frame are extracted from the depth map of the 2D image, and the depth values of the extracted pixels are divided and stored for each frame. process; (B) generating parallax information using the generated depth map; And (c) generating final coefficients of stability (CF) by linear combination by obtaining seven factor values using the depth information and parallax information.

Quality analysis method of a three-dimensional stereoscopic image comprising a.
The method of claim 20,

In the step (c), the three-dimensional stereoscopic image may be obtained by linearly combining the seven coefficient values according to Equation 12 to calculate a final stability coefficient value CF for one frame. Quality analysis method.

[Equation 12]

Here, ω1 to ω7 represent weight values for seven coefficient values.
The method of claim 21,

Step (c) is the same process of calculating the stability coefficient value CF for one frame, and calculates seven coefficient values according to Equation 10 for each frame for every frame, and calculates each frame for every frame. The total stability factor (TCF) is arranged by arranging the seven coefficient values of the frame according to the following equation (13) (Total Compatibility Factor; TCF), and the weights of the seven coefficient values for each frame according to the following equation (14) and a weighting factor (X) value arranged from (ω) values (ω1 to ω7).

[Equation 13]

[Equation 14]

Here, X represents a weighting coefficient value consisting of a 7x1 matrix, S represents a matrix vector for Nx1 subjective evaluation, and N represents the number of frames.
The method of claim 14,

In the step (d), the depth values of each pixel per frame are extracted from the depth map of the 2D image, and the depth values of each extracted pixel are divided and stored for each frame. process; (B) generating parallax information using the generated depth map; And generating error coefficient values for depth reversal of processing an object as a background or processing a background as an object by using a depth value for each frame for the 2D image or using information other than the depth information. (C) process of doing;

Quality analysis method of a three-dimensional stereoscopic image comprising a.
The method of claim 23,

In the step (c), the depth information for each frame is divided into a background part and an object part using a depth value, and the background is converted into an object using information other than depth in the separated background part. A quality analysis method for a 3D stereoscopic image, characterized by detecting a depth reversal to be processed and calculating an error coefficient value.
The method of claim 23,

In the step (c), the difference between the previous frame and the current frame of the 2D image is different from the depth. The method of claim 3, wherein the error coefficient value is calculated.
The method of claim 23,

In the step (c), an error may be detected using a motion vector with information other than the depth, and a motion vector of a background is set in units of blocks, and the motion vectors are units of blocks. And calculating an error coefficient value when the calculated motion vector value differs from the background motion vector by a predetermined amount as an object, and calculates an error coefficient value.
A computer readable medium having recorded thereon a program for executing the quality analysis method of the three-dimensional stereoscopic image according to claim 14.