KR101327709B1 - Apparatus for monitoring video quality and method thereof - Google Patents

Abstract

An apparatus and method for measuring video quality based on user perception are disclosed. The apparatus for measuring video quality collects frame information about a damaged frame and calculates a structural similarity index indicating structural similarity between the damaged frame and the original frame of the damaged frame, and reflects the subjective factors in the structural similarity index. And a user cognitive quality estimator for calculating a video defect index by applying weights. Through this, the subjective factor according to the user's cognition may be reflected in the structural similarity index which is an objective image quality measurement index.

Description

Apparatus and method for measuring video quality {APPARATUS FOR MONITORING VIDEO QUALITY AND METHOD THEREOF}

The present invention relates to measurement of video quality, and more particularly, to an apparatus and method for measuring video quality based on user perception.

The video quality measurement method is a core technology necessary for verifying the performance of video codec, developing new compression coding technology, or evaluating transmission quality. It is utilized.

In general, objective image quality measurement methods can be classified into full-reference (FR), reduced-reference (RR), and no-reference (NR) according to the degree of use of the video image source. The FR uses the complete video source on the sender and the entire corrupted video on the receiver for precise measurement of video quality. The RR measures the image quality with the image represented by a specific value, not all of the video source of the sender. The NR measures the picture quality with only the corrupted video at the receiving end without a video source.

SSIM (Structural Similarity) is an indicator that compares the structural similarity of two images and is used as an indicator to measure the objective quality of the video.Peak Signal-to-Noise Ratio (PSNR) and Mean Squared Error (MSE) are also objective It is widely used as an indicator for measuring quality.

In addition, the Video Quality Model (VQM) is calculated by linear combination of seven parameter values, and the VQM value determined as a value between 0 and 1 may show better image quality as it approaches zero. However, VQM is not suitable for measuring the quality of real-time video due to the high computational complexity.

In addition, since the method of measuring the objective image quality does not reflect the subjective factor, there is a limit in accurately measuring the video quality perceived by the user.

Mean Opinion Score (MOS) is used as an indicator to measure user's empirical quality of video. The quality of experience (QoE), which represents a measure of the quality felt by the user, may be used to quantify and display the MOS, which is the quality of the user. In addition, various methods for estimating the MOS value have been proposed, and in particular, the MOS value can be obtained through a user questionnaire in various network environments and application environments. However, since MOS is measured through a questionnaire for a plurality of users, there is a problem that a lot of cost and time are consumed.

An object of the present invention for solving the above problems is to provide a video quality measuring apparatus that can reflect the quality of the video perceived by the user in the objective quality measurement method.

Another object of the present invention is to provide a video quality measurement method which reduces the complexity of the calculation in calculating the structural similarity between the original frame and the damaged frame.

In order to achieve the above object, a video quality measurement apparatus according to an aspect of the present invention collects frame information about a damaged frame and calculates a structural similarity index indicating structural similarity between the damaged frame and the original frame of the damaged frame. The estimator includes a user perceived quality estimator that calculates a video defect index by applying a weight reflecting a subjective element to the structural similarity index.

Here, the structural similarity estimator generates an error association tree using frame information, and the damaged frame using the structural similarity between the damaged slice constituting the damaged frame and the reference slice referenced by the damaged slice based on the error association tree. And structural similarity index indicating structural similarity between the original frame and the damaged frame can be calculated.

Here, the structure similarity estimator collects frame information including at least one of a frame identifier, a slice identifier, a type of frame, a degree of damage, and a damage position to generate an error association tree based on the type of a frame to which the damaged slice belongs. The performance indicator collection unit may include a structure analysis unit configured to select a reference slice referenced by the damaged slice based on an error association tree and calculate a structure similarity index.

In this case, the subjective factor may reflect the influence of the brightness of the image, the complexity of the frame, and the transition of the scene on the video quality.

Here, the user cognitive quality estimating unit may calculate a weight to reflect a subjective factor and apply the filter unit to the structural similarity index, and apply a logistic regression function to the structural similarity index to which the weight is applied. It may include an indicator converter for calculating a video defect indicator reflecting the shape.

Here, the filter unit may include a luma adjustment filter unit that calculates a luma weight that reflects an effect of the luminance of the target frame on the video quality perceived by the user, and the effect of the complexity of the target frame on the video quality perceived by the user. It may include a frame complexity filter unit for calculating the complexity weight reflecting the, and a scene transition filter unit for calculating the conversion weight reflecting the effect of the conversion of the target frame on the video quality perceived by the user.

Here, the index converter may detect a section of the video where the damage occurs using the structural similarity index, and calculate a video defect index reflecting the shape of the damage based on the change in the structural similarity index.

According to an aspect of the present invention, there is provided a method for measuring video quality, the method comprising: collecting frame information about a corrupted frame and generating an error association tree using the frame information; Calculating a structural similarity index indicating structural similarity between the damaged frame and the original frame of the damaged frame by using the structural similarity between the damaged slice constituting the damaged slice and the reference slice referenced by the damaged slice. And a user perceived quality estimation step of calculating a video defect index by applying the reflected weight.

According to the video quality measuring apparatus and method according to the present invention as described above, the subjective factor according to the user's recognition may be reflected in the structural similarity index which is an objective quality measurement index.

In addition, there is an advantage that the structural similarity between the original frame and the damaged frame can be calculated using the structural similarity between the damaged slice and the reference slice thereof.

1 is a block diagram showing a configuration of a video quality measuring apparatus according to an embodiment of the present invention.
2 is a conceptual diagram illustrating an error association tree according to an embodiment of the present invention.
FIG. 3 is a block diagram illustrating in more detail the configuration of the structural similarity estimator and the user cognitive quality estimator shown in FIG. 1.
4 is a flowchart illustrating a procedure of a method for measuring video quality according to an embodiment of the present invention.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. Like reference numerals are used for like elements in describing each drawing.

When a component is referred to as being "connected" or "connected" to another component, it may be directly connected to or connected to that other component, but it may be understood that other components may be present in between. Should be. 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.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries are to be interpreted as having a meaning consistent with the contextual meaning of the related art and are to be interpreted as either ideal or overly formal in the sense of the present application Do not.

As used herein, the term "frame" may be replaced with another term having an equivalent meaning such as an image, a picture, and the like. Also, a moving picture or a video may be composed of a series of pictures, and each picture may be divided into predetermined regions such as a slice or a block.

In the MPEG standard among the international standards organizations, there are three types of intra coded frames (I frames), predictive coded frames (P frames), and bi-directionally-predicitve coded frames (B frames). A predictive coding method is used.

I-frames, described below, can be encoded using neighboring pixel values within the screen without reference to other frames. The P frame may be encoded with reference to past I frames or P frames. In addition, the B frame may perform prediction by referring to not only I frames and P frames in the past but also future I frames and P frames.

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

1 is a block diagram showing a configuration of a video quality measuring apparatus according to an embodiment of the present invention.

Referring to FIG. 1, an apparatus for measuring video quality according to an embodiment of the present invention includes a structure similarity estimator 100, a user perceived quality estimator 200, and a subjective image quality estimator 300.

Hereinafter, in the description of the present invention, the structural similarity estimator 100, the user cognitive quality estimator 200, and the subjective image quality estimator 300 are disclosed as independent parts, but the structural similarity estimator 100 and the user cognitive quality estimator are described. The government 200 and the subjective image quality estimation unit 300 may be implemented in one single form, one physical device, or one module. In addition, the structural similarity estimator 100, the user cognitive quality estimator 200, and the subjective image quality estimator 300 may be implemented as a plurality of physical devices or groups instead of one physical device or group. .

The structural similarity estimator 100 may collect frame information about a damaged frame and generate an error association tree using the frame information. In addition, the structural similarity estimator 100 uses the structural similarity between the damaged slice constituting the damaged frame and the reference slice referenced by the damaged slice, based on the error association tree. A structural similarity index indicating a can be calculated.

The frame information includes information on a frame identifier, a slice identifier, a type of a frame, and a state (damage degree, damage position) of the frame, and may be collected by the structure similarity estimator 100.

The error association tree indicates an association relationship between a damaged slice and a reference slice thereof, and may be generated according to the type of a frame including the damaged slice. Here, the type of frame may be classified into an I frame, a P frame, and a B frame, and a reference frame referenced by the frame may be determined according to the type of the frame.

Accordingly, the structure similarity estimator 100 may select a reference slice referenced by a damaged slice based on an error association tree, and calculate a structure similarity index using the selected reference slice. Here, the structural similarity index (or “SSIM (Structural Similarity) estimated value”) refers to an index indicating a degree of similarity between a damaged frame and an original frame thereof and may be expressed as a value of 0 to 1. In addition, the closer the structural similarity index is to 1, the higher the similarity between the damaged frame and the original frame thereof.

The structural similarity index between the damaged frame and the original frame of the damaged frame calculated by the structural similarity estimator 100 may be used as an objective index indicating the quality of the video. However, since the structural similarity index has a limitation in reflecting the subjective quality of the video perceived by the user, it is necessary to reflect the subjective element in the structural similarity index.

The user cognitive quality estimation unit 200 calculates a video defect index by reflecting a subjective element in the structural similarity index. The quality of the video may be influenced by subjective factors such as image brightness (brightness), frame complexity, scene transitions, and the like.

The user cognitive quality estimator 200 may calculate and apply a weight to reflect the subjective element in the structural similarity index. The weight may reflect the effect of the luminance of the target frame, the effect of complexity, the effect of switching frames, and the like. That is, the user cognitive quality estimation unit 200 may apply a weight to the structural similarity index.

In addition, the user cognitive quality estimation unit 200 may apply a logistic regression function to the weighted structural similarity index to calculate a video defect index including information about the degree of damage and the type of damage of the video. Can be.

The subjective image quality estimator 300 may calculate a subjective quality index which may be expressed as a Mean Opinion Score (MOS) using a parameter that may be designated for each user in the video defect index. Parameters that can be specified for each user may include information about user patience, service expectations, and the like.

2 is a conceptual diagram illustrating an error association tree according to an embodiment of the present invention.

Referring to FIG. 2, the video is composed of a series of frames, and the frame may be composed of at least one slice. The numbers given to each frame exemplarily indicate the order of successive frames. For example, the 62nd I frame represents a frame preceding the 63rd P frame.

In addition, each node to which the directional connection line is connected represents a slice and may include information on a frame identifier, a slice identifier, a type of frame, and a state of a frame. Here, the black node represents a damaged slice, and the directional connecting line may represent an association relationship (reference relationship) between the slices.

The type of frame may be represented as an I frame, a P frame, or a B frame, and a reference frame referred to for restoring or decoding the frame may be determined according to the type of the frame. For example, a slice 2 belonging to I frame 62 may refer to adjacent slices 1 and 3 without reference to another frame. A slice belonging to the P frame 63 may be restored by utilizing an I frame 62 which is a frame older than the P frame 63. In addition, the B frame 65 may use the P frame 63 and the P frame 66, which are frames of the past and the future, as reference frames. However, the B frame may not be a reference frame for another frame.

Therefore, slices belonging to a damaged P frame or B frame can be restored using their reference slices, and slices belonging to a damaged I frame can be restored using adjacent slices of the same frame.

FIG. 3 is a block diagram illustrating in more detail the configuration of the structural similarity estimator and the user cognitive quality estimator shown in FIG. 1.

Referring to FIG. 3, the structure similarity estimator 100 includes a performance index collector 110 and a structure analyzer 120. In addition, the user cognitive quality estimation unit 200 may include a filter unit 210 and an indicator transformation unit 220.

The performance indicator collecting unit 110 may collect frame information, generate an error association tree based on the information, and transmit the generated frame information to the structure analyzing unit 120. In particular, the performance indicator collection unit 110 according to an embodiment of the present invention may selectively collect frame information on a damaged frame among all the frames, and generate an error association tree using the frame information.

The structure analyzer 120 may calculate a structural similarity index between the damaged frame and the original frame thereof based on the error association tree. That is, the structure analyzer 120 may calculate a structure similarity index based on the type of the frame.

The quality of the video is influenced by the slice reconstruction and error propagation. Since the damaged slice is reconstructed using the reference slice, the quality of the recovered frame may be determined according to the structural similarity of the damaged slice and the reference slice thereto. Therefore, the structure analyzer 120 may estimate a structural similarity index indicating structural similarity between the damaged frame and the original frame using the structural similarity between the damaged slice and the reference slice.

First, the structural similarity index between the original I frame and the damaged frame I 'frame may be calculated using Equation 1 below.

In Equation 1,

eSSIM (I, I ') represents a structural similarity index indicating structural similarity between the original I frame and the damaged I' frame. Here, if eSSIM (I, I ') = 1, there is no damage to the I frame. Q ₃ is a constant for reflecting an error on the reconstruction in the decoding algorithm, and may be set to 1.55 by experiment. S _n represents the number of slices constituting the frame, and ms (missing slice) represents the number of damaged slices.

In addition, SSIM _ms (I, I _r ) represents the structural similarity between the damaged I slice belonging to the damaged I frame and the reference slice I _r slice, and eSSIM (I _r , I _r ') is the original reference frame. It shows the structural similarity between an I _r frame and an I _r 'frame, which is a corrupted reference frame. Here, when eSSIM (I _r , I _r ') = 1, there is no damage to the I _r frame.

Therefore, the structure analysis unit 120 according to an embodiment of the present invention uses the structural similarity between the damaged I slice and the reference slice I _r slice to determine the structure between the original I frame and the damaged I frame. The structural similarity index indicating the similarity can be calculated.

Next, the structural similarity index between the original P frame and the damaged P 'frame may be calculated using Equation 2 below.

In Equation 2,

eSSIM (P, P ') indicates a structural similarity index indicating structural similarity between the original P frame and the damaged P' frame. Here, when eSSIM (P, P ') = 1, there is no damage to the P frame. Q ₂ is a constant for reflecting an error on reconstruction in the decoding algorithm, and may be set to 1.1 by experiment. φ represents the number of intact slices, φ = S _n − Can be found in ms.

SSIM _ms (P, P _r ) represents the structural similarity between the damaged P slice belonging to the damaged P frame and the reference slice P _r slice, and eSSIM (P _r , P _r ') denotes the original reference frame P _r. The structural similarity between the frame and the corrupted reference frame P _r 'frame. Here, when eSSIM (P _r , P _r ') = 1, there is no damage to the P _r frame.

In addition, [SSIM (P, P _r )] is applied to a case where a P frame and a P _r frame which is a reference frame thereof have a long-range dependent relationship. The long-distance dependency relationship means that the frame referred to by the lost reference frame is lost because the reference frame is lost. For example, referring to FIG. 2, when the 66th P frame 66 is lost and the 69th P frame 69 refers to the 63rd P frame 63, it may be regarded as having a long distance dependency relationship. That is, [SSIM (P, P _r)] is a reference in a case where a P frame and a reference frame is a P _r frames thereof in the long-distance-dependent (long-range dependent) relationships, damaged P slices and for it belongs to the P-frame The structural similarity between slices of P _r slices is shown.

Therefore, the structure analysis unit 120 according to an embodiment of the present invention uses the structural similarity between the damaged P slice and the reference frame P _r slice thereof to determine the structure between the original P frame and the damaged P 'frame. The structural similarity index indicating the similarity can be calculated.

Next, the structural similarity index between the original B frame and the damaged B 'frame may be calculated using Equation 3 below.

In Equation 3,

eSSIM (B, B ') represents a structural similarity index indicating structural similarity between the original B frame and the damaged frame B'. Here, if eSSIM (B, B ') = 1, there is no damage to the B frame.

SSIM _ms (B, B _r ) represents the structural similarity between the damaged B slice belonging to the damaged B frame and the reference slice B _r slice, and eSSIM (B _r , B _r ') denotes the original reference frame B _r. The structural similarity between the frame and the damaged reference frame B _r 'frame. Here, when eSSIM (B _r , B _r ') = 1, it indicates that there is no damage to the B _r frame. In addition, in the case of the B frame, it is possible to refer to the frame of the future as well as the past, where B _r1 frame (slice) is a past reference frame (slice) and B _r2 frame (slice) means a future reference frame (slice). can do. However, when the B frame utilizes only one reference frame, the structural similarity index between the original B frame and the damaged B 'frame, which is a damaged frame, may be calculated by applying Equation 2 as in the P frame.

Therefore, the structure analysis unit 120 according to the embodiment of the present invention uses the structural similarity between the damaged B slice and the reference slice B _r slice thereof, and the structure between the original B frame and the damaged frame B ′ frame thereof. The structural similarity index indicating the similarity can be calculated.

In addition, referring to FIG. 3, the user cognitive quality estimator 200 includes a filter 210 and an index converter 220.

The filter unit 210 may calculate a weight to reflect a subjective factor that affects the user's perception of the quality of the video and apply it to the structural similarity index. The user may perceive the quality of the video differently according to factors such as luminance (lumina or luma) of the image, complexity of the frame, transition of the scene, and the like. For example, it is difficult for a user (a viewer) to notice a slight degradation of video quality in dark scenes and a minor degradation of video quality in fast transition scenes. In addition, it may be difficult for a user to notice minor degradation of video quality in a complex image.

The filter unit 210 may include a luma adjust filter 211, a frame complexity filter 212, and a scene change filter 213.

The luma adjustment filter unit 211 is based on the fact that the perception of human video quality is affected by the luminance of the object. That is, the degradation of video quality in bright scenes may have a greater impact on the user than the degradation of video quality in dark scenes. The structural similarity index may be calculated for a series of frames, and the luma adjustment filter 211 may apply a weight to the structural similarity index that reflects the effect of the brightness of the image on the video quality perceived by the user. . That is, the weight that reflects the effect of the luminance of the target frame on the video quality perceived by the user may be referred to as luma weight.

The luma weight may be calculated using Equation 4 below.

In Equation 4,

는 x프레임의 휘도를

는 y프레임의 휘도를 나타낸다.

는 휘도에 따라 정해지는 루마 가중치이다. x and y represent the two frames to compare,

Is the luminance of the x frame

Denotes the luminance of the y frame.

Is a luma weight determined by luminance.

If any one of the luminance of the x frame and the y frame is greater than 50, the luma weight is set to one. In addition, when the luminance of both the x frame and the y frame is less than 50, and the average of the luminance of the x frame and the y frame is less than 40, the luma weight may be set to zero. Also, if the luminance of both x and y frames is less than 50 and the average of the luminance of x and y frames is greater than 40 and is equal to or less than 50, subtract 40 from the average of the luminance of x and y frames. The value divided by 10 is used as luma weight. That is, it can be seen that as the luminance of the frame increases, the value of the luma weight increases. The luma weights calculated by the luma control filter unit 211 may be transmitted to the structure analyzer 120, and the structure analyzer 120 receiving the luma weights may be output by applying luma weights to the structural similarity index. .

The frame complexity filter unit 212 is based on the fact that human perception of video quality is affected by the complexity of the object. That is, the effect of the degradation of video quality on the user may vary depending on the complexity of the frame.

The frame complexity filter unit 212 calculates the complexity weights to reflect these characteristics and applies them to the structural similarity index so as to obtain a more accurate structural similarity index reflecting the video quality actually recognized by the user. That is, the weight reflecting the influence of the complexity of the target frame on the video quality perceived by the user may be referred to as the complexity weight.

First, Mean Absolute Difference (MAD) may be used as a method for calculating a frame complexity (FC) indicating the complexity of a frame. However, the frame complexity filter 212 according to an embodiment of the present invention may modify the MAD. FC can be calculated using Equation 5.

In Equation 5,

은 샘플 프레임 r내의 픽셀k의 휘도를 나타내고,

은 샘플 프레임 r내의 모든 픽셀의 휘도의 평균값을 나타낸다. R means the number of samples in one frame, and one sample is composed of m × n pixels. Also,

Denotes the luminance of pixel k in sample frame r,

Denotes an average value of luminance of all pixels in the sample frame r.

)는 다음의 수학식 6를 이용하여 구조 유사도 지표에 적용될 수 있다. Complexity weights (

) May be applied to the structural similarity index using Equation 6 below.

In Equation 6,

)가 결정되고, 결정된 복잡도 가중치(

)는 구조 유사도 지표가 0.60보다 큰 경우에만 적용될 수 있다. Based on the complexity (FC) indicator of the complexity of the frame, the complexity weight (

) Is determined, and the determined complexity weight (

) Can be applied only if the structural similarity index is greater than 0.60.

는 FC가 25보다 작거나 같은 경우에 1이 되고, FC가 30보다 크거나 같은 경우에는 0이 될 수 있다. 또한, FC가 25와 30사이에 있는 경우,

는 0과 1사이의 값을 가질 수 있다. 즉, 프레임의 복잡도가 낮을수록 복잡도 가중치의 값이 커질 수 있다.
for example,

May be 1 if the FC is less than or equal to 25, and may be 0 if the FC is greater than or equal to 30. Also, if the FC is between 25 and 30,

May have a value between 0 and 1. That is, the lower the complexity of the frame, the larger the value of the complexity weight.

The scene change filter unit 213 is based on the fact that the perception of the video quality of the human being is affected by the conversion of the object. In other words, humans may not feel the deterioration of video quality at the moment when the scene changes, and this effect of scene change is known to last about 1/3 second. Detection of scene transitions can be achieved through color changes in two consecutive frames. In addition, a method using a block-based color histogram may be used to detect a scene change.

A scene change detection index (SC) for detecting a scene change may be calculated using Equation 7 below.

In Equation 7,

f _i may represent the i-th frame, and f _{i -1} may represent a previous frame of f _i . b may mean one block when the frame is divided into 16 blocks, and c may mean one of R, G, and B components. In addition, k means a value that each of the R, G, and B components can have, the maximum value can be L ^C. Here, L ^C means the maximum value of each R, G, B component, and mn may represent the size of the frame.

In addition, H represents a histogram function, and H (f _i , c, b, k)-H (f _{i -1} , c, b, k) indicates the difference between successive frames for each c, b, k value. Can be represented. For example, when c = R, b = 1, and k = 10, the difference in the number of pixels having an R value of 10 among the pixels in block 1 in a subsequent frame is calculated.

Therefore, after dividing a frame into 16 blocks, color histograms for each block may be generated for R, G, and B color elements. Through this, the scene change detection index SC may be calculated using the sum of the differences in the color histograms of the blocks belonging to two consecutive frames.

)는 다음의 수학식 8을 이용하여 구조 유사도 지표에 적용될 수 있다. Conversion weight (

) May be applied to the structural similarity index using Equation 8 below.

In Equation 8,

)가 결정되고, 결정된 전환 가중치(

)는 구조 유사도 지표에 가중치로 적용될 수 있다. Based on cutaway detection metrics (SC)

) Is determined, and the determined conversion weight (

) Can be applied as a weight to the structural similarity index.

First, condition A refers to a case in which the switching of the frame occurs within 1/6 second, and condition B refers to a case in which the switching of the frame occurs after 1/6 second.

는 0이 되고, SC가 1.7보다 크거나 같고 3.0보다 작으며 조건B에 해당하는 경우에

는 0이 될 수 있으며, 그 외의 경우에 있어서

는 1이 될 수 있다.
For example, if SC is greater than or equal to 3.0 and corresponds to condition A

Is 0 and SC is greater than or equal to 1.7, less than 3.0, and corresponds to condition B.

Can be 0, in other cases

Can be 1.

The index transformation unit 220 may calculate a video defect index by applying a logistic regression function to the weighted structural similarity index. The indicator converter 220 may detect a section of the video in which the damage occurs by using the structural similarity index, and calculate a video defect index that reflects the shape of the damage based on the change in the structural similarity index.

The video defect indicator may reflect the type of damage represented by distortion, freezing, discontinuity, stutter, and glitch.

The structural similarity index may be calculated for each frame, and the video may have a series of structural similarity indexes. When the structural similarity index is 1, there is no damage. Therefore, the structural similarity index may be classified by breaking the section in which the continuous similarity index does not appear. Here, the damage event refers to the section of the video where the damage occurs, and the damage event is classified into distortion, freezing, discounter, stirrer, and glitch based on the change in the structural similarity index. Can be.

Distortion refers to a case in which a user perceptible distortion occurs in a frame in a damage event, and freezing refers to a series of identical frames. In addition, the discontinuity means that frames that are not supposed to be played back are displayed in succession, and stutter means that freezing occurs less than a predetermined time (0.5 seconds). , Glitch may mean a case in which the distortion occurs for less than a predetermined time (0.5 seconds).

4 is a flowchart illustrating a procedure of a method for measuring video quality according to an embodiment of the present invention.

Referring to FIG. 4, a video quality measurement method according to an embodiment of the present invention includes generating an error association tree, calculating a structural similarity index, and estimating user perceived quality.

The video quality measurement method according to an embodiment of the present invention may be implemented as a plurality of physical devices or groups instead of one physical device or group, and each step included in the video quality measurement method according to the embodiment of the present invention It can be implemented by one physical device, one module or software.

For example, referring to FIG. 3, the generating of the error association tree (S200) may be performed by the performance indicator collecting unit 110, and calculating the structural similarity index (S300) may be performed by the structure analyzing unit 120. The user cognitive quality estimating step S500 may be performed by the user cognitive quality estimating unit 200.

Generating an error association tree (S200) may collect frame information on the corrupted frame (S100), and generate an error association tree using the collected frame information. Here, the error association tree may indicate an association relationship between a damaged slice and a reference slice thereof. That is, the error association tree may be generated based on the type of the frame to which the damaged slice belongs by collecting frame information including at least one of a frame identifier, a slice identifier, a type of frame, a degree of damage, and a location of damage.

Computing the structural similarity index (S300) is a structure between the damaged frame and the reference frame referenced by the damaged frame using the structural similarity between the damaged slice constituting the damaged frame and the reference slice referenced by the damaged slice based on the error association tree. Similarity indicators can be calculated.

In the user perceived quality estimation step S500, a video defect index may be calculated by applying a weight reflecting a subjective element to the structural similarity index S400.

Therefore, the user perception quality estimation step may be applied to the structural similarity index by calculating a weight for reflecting a subjective element including information about the brightness of the image, the frame complexity, and the change of the scene.

In addition, in the user cognition quality estimation step, a logistic regression function may be applied to the structural similarity index to which the weight is applied to calculate a video defect index reflecting the shape of the damage. That is, by using the structural similarity index, the section of the video where the damage occurs can be detected, and the video defect index reflecting the shape of the damage can be calculated based on the change in the structural similarity index. Here, the form of damage may be represented by distortion, freezing, discontinuity, stutter, and glitch.

When using the apparatus and method for measuring video quality according to the present invention described above, a structural similarity index indicating structural similarity between an original frame and a damaged frame may be calculated using structural similarity between a damaged slice and a reference slice thereof. Can be. In addition, by applying a weight reflecting the subjective factor to the structural similarity index, it is possible to measure the video quality perceived by the user.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the present invention as defined by the following claims It can be understood that

100: structural similarity estimation unit 110: performance indicator collection unit
120: structural analysis unit
200: user cognitive quality estimation unit 210: filter unit
211: luma adjustment filter unit 212: frame complexity filter unit
213: Scene transition filter unit 220: Indicator conversion unit
300: subjective image quality estimation unit

Claims (13)

Collect frame information about the corrupted frame and generate an error association tree using the frame information, and structural similarity between the damaged slice constituting the damaged frame and the reference slice referenced by the damaged slice based on the error association tree A structural similarity estimator for calculating a structural similarity index indicating structural similarity between the damaged frame and the original frame of the damaged frame by using a? And
And a user cognitive quality estimator for calculating a video defect index by applying a weight reflecting a subjective element to the structural similarity index. delete The method of claim 1, wherein the structural similarity estimator,
A performance indicator collection unit configured to collect the frame information including at least one of a frame identifier, a slice identifier, a type of frame, a degree of damage, and a damage location to generate the error association tree based on the type of the frame to which the damaged slice belongs; ; And
And a structure analyzer configured to calculate the structure similarity index by selecting a reference slice referenced by the damaged slice based on the error association tree. The method according to claim 3,
The type of frame includes I frame, P frame, B frame,
And the error association tree indicates an association relationship between the damaged slice and the reference slice according to the type of the frame. The method according to claim 1,
The subjective element is a video quality measuring apparatus, characterized in that reflects the effect of the brightness of the image, the complexity of the frame, the scene change on the video quality. The method of claim 1, wherein the user cognitive quality estimation unit,
A filter unit calculating a weight for reflecting the subjective element and applying the same to the structural similarity index; And
And an index converter configured to apply a logistic regression function to the structural similarity index to which the weight is applied, to calculate a video defect index reflecting the shape of the damage. The method according to claim 6, The filter unit,
A luma adjustment filter for calculating a luma weight reflecting an effect of the luminance of the target frame on the video quality perceived by the user;
A frame complexity filter unit for calculating a complexity weight that reflects an effect of the complexity of the target frame on the video quality perceived by the user; And
And a scene change filter configured to calculate a conversion weight that reflects the effect of the conversion of the target frame on the video quality perceived by the user. The method according to claim 6, wherein the indicator conversion unit,
And detecting a section of a video in which damage occurs using the structural similarity index, and calculating a video defect index reflecting the shape of the damage based on a change in the structural similarity index. The method of claim 6, wherein the video defect indicator,
A device for measuring video quality that reflects the type of damage represented by distortion, freezing, discontinuity, stutter, and glitch. In measuring video quality,
Collecting frame information on the corrupted frame and generating an error association tree using the frame information;
A structural similarity index indicating structural similarity between the damaged frame and the original frame of the damaged frame using the structural similarity between the damaged slice constituting the damaged frame and the reference slice referenced by the damaged slice based on the error association tree. Calculating; And
And a user perceived quality estimation step of calculating a video defect index by applying a weight reflecting a subjective element to the structural similarity index. The method of claim 10, wherein generating the error association tree,
Generating the error association tree based on the type of the frame to which the damaged slice belongs by collecting the frame information including at least one of a frame identifier, a slice identifier, a type of frame, a degree of damage, and a damage location. How to measure video quality. The method of claim 10, wherein the user perceived quality estimation step,
Calculating weights for reflecting subjective elements including information on brightness, frame complexity, and transition of an image, and applying the same to the structural similarity index; And
And applying a logistic regression function to the structural similarity index to which the weight is applied, to calculate a video defect index reflecting the shape of the damage. The method of claim 12, wherein calculating the video defect indicator comprises:
And detecting a section of a video in which damage occurs using the structural similarity index, and calculating a video defect index reflecting the shape of the damage based on a change in the structural similarity index.

Images