KR100893609B1 - Apparatus and Method by Using Human Visual Characteristics - Google Patents

Abstract

The present invention relates to an apparatus and method for measuring image quality using human visual characteristics.

An image quality measuring apparatus according to an exemplary embodiment of the present invention receives a structural image indicator and a structural similarity indicator calculation unit configured to receive a raw image signal and a received image signal, and calculate a numerical value of the structural similarity indicator to determine how similar the received original image signal and the received image signal are. ; A visual weight calculator configured to calculate one or more visual weights by reflecting factors affecting human visual characteristics—information of any one or more of luminance, region of interest, motion, and space-time frequency; And generating final weight information by multiplying the calculated one or more visual weights, and multiplying the calculated structural similarity index by the calculated final weight information to calculate final image quality information.

The present invention can be expected to have an effect of obtaining a result similar to an evaluation score, which is actually directly scored by a person by reflecting human visual characteristics in measuring image quality of multimedia.

Human vision, weights, image quality, subjective, objective

Description

Apparatus and Method by Using Human Visual Characteristics}

1 is a block diagram schematically illustrating an internal configuration of an image quality measuring apparatus according to an exemplary embodiment of the present invention.

FIG. 2 is a diagram illustrating variables in an algorithm for calculating a weight based on a region of interest according to an exemplary embodiment of the present invention.

3 is a view showing the structure of the eye to explain a region of interest according to an embodiment of the present invention.

4 is a view for explaining a region of interest in human vision according to an embodiment of the present invention.

FIG. 5 is a diagram illustrating another example of a variable including an equation in FIG. 2 illustrating a variable in an algorithm for calculating a weight according to an ROI according to an exemplary embodiment of the present invention.

6 is a view for explaining the degree of sensitivity of human vision according to the space-time frequency according to an embodiment of the present invention.

7 is a view for explaining a method for measuring image quality according to an embodiment of the present invention.

The present invention relates to an apparatus and method for measuring image quality using human visual characteristics, and more particularly, to an apparatus and method for measuring image quality using visual weights according to human visual characteristics.

In recent years, the importance of video quality is emerging for the success of video services. The method of evaluating the quality of an image can be largely divided into a subjective method and an objective method. Subjective image quality evaluation is a method of determining the quality level by voting by multiple evaluators. Even if the evaluation targets are the same, the results may change when the evaluation conditions are different. Therefore, the statistical values of the multiple experimental results are used, and the time and financial burden for measurement is reduced. Big.

In addition, subjective image quality evaluation may be considered to reflect human vision well, but there is a problem that the value varies according to the evaluator, and the value may change every time. The most representative subjective quality assessment method is MOS (Mean Opinion Score).

Objective image quality evaluation is a method of deriving the quality level by mathematical calculation. Since there is an accurate calculation method, it is possible to expect a certain result regardless of the evaluator or the evaluation condition if the evaluation target is the same. The most representative objective quality evaluation methods are PSNR (Peak Signal to Noise Ratio) and MSE (Mean Square Error), which are methods for directly comparing the difference between each pixel value of the original image and the degraded image.

However, objective image quality evaluation does not reflect human visual characteristics properly, and images with lower PSNR values may look better to the human eye. Image quality measurement technology reflecting human visual characteristics is an internationally unexplored field, and since standardization is currently in progress, it is urgent to develop related technologies and participate in standardization.

Current video quality measurement methods lack the technology and equipment to accurately reflect the image quality felt by humans, and in particular, it is very difficult to measure the video quality of various types of video terminals such as video phones, PDAs, and PC soft phones. to be.

In order to solve such a problem, the present invention is to provide an apparatus and method for measuring image quality using visual weights according to human visual characteristics.

According to an aspect of the present invention, an image quality measuring apparatus receives an original image signal and a received image signal, and digitizes structural similarity indicators of how similar the original image signal and the received image signal are. Structural similarity index calculation unit to calculate; A visual weight calculator configured to calculate one or more visual weights by reflecting factors affecting human visual characteristics—information of any one or more of luminance, region of interest, motion, and space-time frequency; And an image quality measuring unit configured to generate final weight information by multiplying the calculated one or more visual weights by each other, and multiplying the generated final weight information by the calculated structural similarity index to calculate final image quality information.

According to an aspect of the present invention, there is provided a method for measuring image quality (a) receiving an original image signal and a received image signal from an external memory, calculating a structural similarity index, and calculating a human visual image based on the received original image signal. Calculating a plurality of visual weight information reflecting factors influencing the characteristic—information of any one or more of luminance, region of interest, motion, and space-time frequency; And (b) generating final weight information by multiplying the calculated plurality of visual weight information with each other, and calculating final image quality information by multiplying the calculated structural similarity index by the generated final weight information.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention. As those skilled in the art would realize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the present invention. In the drawings, parts irrelevant to the description are omitted in order to clearly describe the present invention, and like reference numerals designate like parts throughout the specification.

Throughout the specification, when a part is said to "include" a certain component, it means that it can further include other components, without excluding other components unless specifically stated otherwise. In addition, the terms “… unit”, “… unit”, “module”, “block”, etc. described in the specification mean a unit that processes at least one function or operation, which is hardware or software or a combination of hardware and software. It can be implemented as.

Now, an apparatus and method for measuring image quality using human visual characteristics according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram schematically illustrating an internal configuration of an image quality measuring apparatus 100 according to an exemplary embodiment of the present invention.

The image quality measuring apparatus 100 according to an exemplary embodiment of the present invention calculates a similar degree between the original image and the received image, and then weights the image to reflect the human visual characteristics. That is, even if the pixel values of the original image and the received image are similar, the image quality measuring apparatus 100 may feel the image quality according to the image quality that the human eye feels, the object of interest in the screen, the degree of contrast, frequency, and motion. Because it is different.

The image quality measuring apparatus 100 according to the exemplary embodiment of the present invention includes a structural similarity index calculator 200, a visual weight calculator 300, and an image quality measurer 400. In addition, it includes an external memory 500 for storing the original image and the received image.

The structural similarity index calculator 200 receives the original image and the decoded received image from the external memory 500, and calculates a degree of similarity between how similar the block of the original image and the block of the received image are. Here, the degree of similarity can be obtained by defining and calculating a structural similarity index measure (SSIM).

The similarity of luminance, the similarity of variance, and the degree of covariance are calculated by using the following [Equation 1] as a component for obtaining structural similarity index.

는 원영상의 평균,

는 수신 영상의 평균,

는 원영상과 수신 영상의 분산(Variance),

는 원영상의 표준 편차,

는 수신 영상의 표준 편차,

는 공분산(Covariance)를 의미한다. 여기서, 평균은 화소값의 평균을 의미한다.Where X is the pixel of the block in the original image, Y is the pixel of the block in the received image,

Is the average of the original image,

Is the average of the received image,

Variance of the original image and the received image,

Is the standard deviation of the original image,

Is the standard deviation of the incoming image,

Means covariance. Here, the mean means an average of pixel values.

Each degree of similarity has a value between 0 and 1, and is 1 in the most similar case, that is, when all pixels of the original image block and the received image block have the same value.

, 분산의 유사 정도

, 공분산 비율 정도

를 곱하여 계산한다. 따라서, 구조적 유사 지표는 다음의 [수학식 2]와 같다.The structural similarity index shows the similarity of luminance

, Similarity of variance

Covariance ratio

Calculate by multiplying by. Therefore, the structural similarity index is shown in Equation 2 below.

라는 조건이 반드시 필요하다. 만일 모든 화소가 0이면

가 성립할 수 있으므로 이를 방지하기 위하여 [수학식 2]에 상수를 더하면 최종 구하고자 하는 구조적 유사 지표는 다음의 [수학식 3]과 같다.At this time,

Condition is necessary. If all pixels are zero

In order to prevent this problem, if you add a constant to [Equation 2], the structural similarity index to be obtained is as shown in [Equation 3].

Where C ₁ and C ₂ are constants.

The visual weight calculator 300 calculates one or more visual weights by reflecting factors affecting human visual characteristics—information of any one or more of luminance, ROI, motion, and space-time frequency.

The visual weight calculator 300 may include an image color signal transform unit (YCbCr Coordinate Transform) 301, a luminance weight calculator 302, a region of interest distance measurement unit 303, and a region of interest weight. Foveal Weight 304, Motion Estimation 305, Motion Weight 306, Channel Decomposition 307 and Space-Time Frequency Weight Calculator (Spatiotemporal CSF Weight) 308.

The image color signal converting unit (YCbCr Coordinate Transform) 301 receives the original image signal from the external memory 500, converts the original image signal into a YCbCr signal when the original image signal is an RGB signal, and if the original image signal is a YCbCr signal, Output the original video signal.

The luminance weight calculator 302 receives a YCbCr signal (ie, a luminance signal + color signal) from the image color signal converter 301, extracts only a luminance signal among the received YCbCr signals, and based on the extracted luminance signal. By using the luminance weighting algorithm (hereinafter, [Equation 4], [Equation 5]) to calculate the weight (W _L ) by the luminance.

The characteristics of human vision seem to be relatively small than changes near bright pixels. Therefore, the change in the pixel value felt by human vision varies depending on the contrast of the background pixel even when the same pixel value is changed. Sensitivity to human vision is sharply lowered on very dark backgrounds.

Therefore, the weight by the luminance was modeled as a power low as shown in Equation 4 below.

Where Y is the first luminance value and c and r are positive constants that determine the shape of the function.

Using the above Equation 4, the weight by the luminance of the K-th block is calculated using the following Equation 5.

Where L is the second luminance value and Lt is the threshold value. Parameter values for multimedia applications are c = 1, r = 1 / 2.2 and Lt = 50.

In other words, when the image quality measurement is for multimedia use, substitute L / Lt of [Equation 5] into Y of [Equation 4], and substitute c = 1 and r = 1 / 2.2 into [Equation 4]. To calculate the weight by luminance.

The region of interest distance measurement unit 303 receives an original image signal from the external memory 500 and performs a perforation point for each block through a face detection algorithm known from the received original image signal. (Foveation Point) is determined, and the distance dx from the foviation point to the middle (X, Y) of each block is generated for each block. Here, the foviation point is determined through the face detection algorithm described above, but may be determined using a mouse, an eye tracker, or the like at the reception side.

When an observer looks at an image, the pixels at all spatial positions of the image are not equally recognized. The point of interest is most accurately perceived, and the further away from the point, the less accurately perceived by the human eye. At this time, as shown in Figs. 2 and 3, the point gazes with interest becomes a poavation point, and the image of the poavation point forms a point where the optic nerve called the fovea of the retina is most distributed. do. Thus, the density of the optic nerve decreases rapidly away from the fovea.

When you move away from the point of staring in the image, the cognitive ability of the person is drastically reduced, and the spatial importance in the image can be obtained by considering this.

The importance of using the spatial characteristics of vision for each pixel can be calculated by considering the distance between the human eye and the image and the distance between the pixels from the foviation point when the posi- tion point is determined.

For example, when making a video call, as shown in FIG. 4, the face portion of the person will be the region of interest. In other words, the highest resolution is in the region of interest to which human vision is focused, and the further the resolution is, the lower the resolution becomes.

When the focal point is determined by the ROI distance measurer 303, the ROI weight calculator 304 obtains the importance using the spatial characteristics by assigning a weight according to the distance from this point. Will be.

를 계산한다.The region of interest weight calculation unit 304 receives a distance dx from the region of interest distance measurement unit 303 to the middle (X, Y) of each block at each posi- tion point and detects the maximum frequency.

Calculate

각도에서의 최대로 감지할 수 있는 주파수

는 다음의 [수학식 6] 및 [수학식 7]를 이용하여 계산한다.

Maximum detectable frequency in degrees

Is calculated using the following [Equation 6] and [Equation 7].

Where i _p (pixel) and i _d (unit of distance) correspond to the size of the image, v _d (unit of distance) is the distance between the observer and the image, and d _x (pixel) is the middle of the block at the focal point ( The distance to X and Y) is shown.

는 포비에이션 포인트로부터의 거리에 따른 공간 주파수의 감쇄를 조절하는 파라미터이다. 멀티미디어 응용에서의 각 상수의 값들은

이다.here,

Is a parameter for adjusting the attenuation of the spatial frequency according to the distance from the fob point. The values of each constant in multimedia applications

to be.

는 변환 인수

(Cycle/Pixel)를 이용하여

(Cycle/Pixel)으로 변환된다. 여기서,

는 도 5에 도시된 바와 같이, 블록당 인간 시각에 들어오는 각도이다. 즉,

(Cycles/Degree)에

(Degree/Blocks)를 곱함으로써

(Cycles/Blocks)로 단위를 바뀌준다.

Is a conversion argument

Using (Cycle / Pixel)

Converted to (Cycle / Pixel). here,

Is the angle entering the human vision per block, as shown in FIG. In other words,

To (Cycles / Degree)

By multiplying (Degree / Blocks)

Change units to (Cycles / Blocks).

는 도 5에 도시된 바와 같이, 다음의 [수학식 8]과 같다.here,

5 is shown in Equation 8 below.

는 최소 주파수

와 최대 정규 주파수 0.5 사이의 값을 갖도록 다음의 [수학식 9]과 [수학식 10]에 의해서 제한된다.after,

Is the minimum frequency

It is limited by the following Equations 9 and 10 to have a value between and the maximum normal frequency 0.5.

는 0.07로 설정한다. Where minimum frequency

Is set to 0.07.

는 0.5보다 크면 0.5로 정하고, 다시 [수학식 9]의 파라미터로 대입하여

을 0.07보다 작으면 0.07로 정함으로써 0.07 ~ 0.5 사이의 값으로 제한한다.Generated by Equation 9 above.

Is greater than 0.5, it is set to 0.5, and is substituted by the parameter of Equation 9 again.

If less than 0.07, set 0.07 to limit the value between 0.07 and 0.5.

Such a perceptible spatial frequency model may indicate a degree of resolution drop according to a distance from the ROI.

The ROI weight calculator 304 receives the dx received from the ROI distance measurer 303, and based on the received dx, the ROI weighting algorithm (Equation 6, Equation 7, Equation 8], [Equation 9], [Equation 10], [Equation 11]) to calculate the weight by the region of interest.

를 계산하고, 계산된

를 [수학식 7]에 대입하여

를 계산한다. 이어서, 관심 영역 가중치 계산부(304)는 계산된

를 [수학식 9]에 대입하여

를 계산하고, 계산된

를 [수학식 10]에 대입하여 최종

을 계산한다.That is, the ROI weight calculation unit 304 substitutes the received dx into Equation 6

Is calculated,

Into Equation 7

Calculate Subsequently, the ROI weight calculator 304 calculates the calculated

By substituting in Equation 9

Is calculated,

Is substituted into [Equation 10]

Calculate

The weight by the ROI is expressed by Equation 11 below.

의 파라미터가 0인 경우, 포비에이션 포인트와 일치하는 곳을 의미한다.here,

If the parameter of 0 is equal to the position of the fovi point.

The motion estimator 305 receives the original video signal from the external memory 500, compares the received original video signal with a previous video signal previously stored in the memory (not shown), and calculates a motion vector for each macro block. Calculate

The motion weight calculation unit 306 may use the motion weighting algorithm 306 based on the motion vector of each macro block calculated from the motion measuring unit 305. Calculate the weight Wm.

In most codecs, each frame is divided into macroblocks (16 × 16), and a motion vector of each macroblock (in which direction) moves through motion estimation with a previous frame. Therefore, since the weight by motion according to the embodiment of the present invention can easily find a rapidly changing part for each region of the frame, it is defined by using the magnitude of the motion vector obtained by motion estimation.

The weight by motion is calculated using the following [Equation 12].

는 k번째 블록의 움직임에 의한 가중치이고, K는 쓰레스홀드(Threshold)이며,

는 최대 움직임 벡터값이다.here,

Is the weight by the movement of the k-th block, K is the threshold,

Is the maximum motion vector value.

는 각각 16과 34이다. 예를 들어, 움직임 벡터는 (24, 24)이면, 크기는 34(

)이다.K in multimedia applications

Are 16 and 34 respectively. For example, if the motion vector is (24, 24), the magnitude is 34 (

)to be.

The spatial frequency converter 307 extracts a frequency component for giving a weight according to the space-time frequency. The spatial frequency converter 307 receives the original image signal and generates a discrete cosine transform (DCT) coefficient for each block (8x8 pixels).

The spatial temporal frequency weight calculation unit 308 receives the DCT coefficients of each block generated by the spatial frequency transform unit 307 and performs the space-time frequency conversion algorithm ([Equation 13], [Equation 14], [ Equation 15]) to calculate the weight (Wc) according to the space-time frequency.

As shown in FIG. 6, the human vision becomes more sensitive as the spatial frequency is lower, reaches a peak at some point, and becomes lower again. Similarly, human vision becomes insensitive as higher and higher frequency, but peaks at some point.

Equation 13 below shows the sensitivity of human vision as a function of time and space frequency.

Contrast Sensitivity Function (CSF) based on space-time frequency is expressed by Equation 13 below.

Here, f _s is a time frequency, f _t is a spatial frequency.

The weight according to the space-time frequency is calculated using Equation 5 below.

는 최정점의 공간 주파수와 시간 주파수를 나타내며, 멀티미디어 용도에서 각각 8cycle/degree와 10 Hz이다.here,

Represents the highest spatial and temporal frequencies and is 8cycle / degree and 10 Hz respectively for multimedia applications.

는 정규화시킨 이산 코사인 변환(Discrete Cosine Transform, 이하 'DCT'라 칭함) 계수이고, P는 DCT 계수를 공간 주파수로 변환하는 변환 지수이다. 여기서, 공간 픽셀의 X, Y와 공간 주파수의 코피션트(Coefficient)의 U, V는 대응된다.here,

Is a normalized Discrete Cosine Transform (DCT) coefficient, and P is a transform index for transforming DCT coefficients into spatial frequency. Here, X and Y of the spatial pixel and U and V of the coefficient of the spatial frequency correspond.

P described above is calculated using Equation 15 below.

(Pixels/Degree)

(Pixels / Degree)

는 뷰잉 디스턴스(Viewing Distance)(Unit of Length)이다.Where r is Display Resolution (Pixels / Unit of Length),

Is the Viewing Distance (Unit of Length).

=12(Inches)라고 하면 p=15.1(Pixels/Degree)이다.For example, r = 72 (Pixels / Inch),

= 12 (Inches), p = 15.1 (Pixels / Degree).

라고 표현할 수 있다.At this time,

Can be expressed.

Since each DCT frequency (u & v) is defined in an 8x8 block, dividing by 16 yields a Maximum Normalized Frequency (Cycles / Pixels) of 0.5.

The image quality measuring unit 400 is a luminance weighting unit, an ROI weighting unit, from the luminance weighting unit 302, the ROI weighting unit 304, the motion weighting unit 306, and the space-time frequency weighting unit 308, respectively. Receive motion weights, space-time frequency weights. Subsequently, the image quality measuring unit 400 calculates final weight information by multiplying the luminance weight, the ROI weight, the motion weight, and the space-time frequency weight.

Final weight information is given by Equation 16 below.

It can be seen that the final weight information consists of the product of each of the four weights, and the subscript k represents the k-th block.

는 각각 시공간 주파수, 휘도, 관심 영역, 움직임에 의한 가중치를 나타낸 것이다. 이때, i, j, l, n는 각각 최적의 가중치를 위한 변수이며, 멀티미디어 용도를 위한 디폴트값은 각각 1이다.bracket

Are weights by space-time frequency, luminance, region of interest, and motion, respectively. In this case, i, j, l, n are variables for the optimal weight, respectively, the default value for the multimedia use is 1 respectively.

The image quality measuring unit 400 receives a structural similarity index generated for each image block from the structural similarity index calculator 200, and multiplies the received structural similarity index by the final weight information to calculate final image quality information.

More specifically, the final image quality information is calculated using Equation 17 below.

는 [수학식 17]에 의한 최종 가중치 정보,

는 각각 b번째 블록의 원영상 신호와 수신 영상 신호이다.Where B is the total number of blocks,

Is the final weight information according to [Equation 17],

Are respectively the original video signal and the received video signal of the b-th block.

Next, an image quality measuring method will be described with reference to FIG. 7.

7 is a view for explaining a method for measuring image quality using human visual characteristics according to an embodiment of the present invention.

Hereinafter, the structural similarity index calculation, the luminance weighting algorithm, the ROI weighting algorithm, the motion weighting algorithm, and the space-time frequency conversion algorithm are described in FIG.

The structural similarity index calculator 200 receives the original image signal and the received image signal from the external memory 500, calculates the structural similarity index, and transmits the structural similarity index to the image quality measuring unit 400 (S100).

When the original image signal received from the external memory 500 is an RGB signal, the image color signal converter 301 converts the image signal into a video signal (YCbCr signal) including a luminance signal and a color signal, and the received original image signal is a YCbCr signal. In this case, the original video signal is transmitted to the luminance weight calculator 302 as it is.

The luminance weight calculator 302 extracts a luminance signal from the original image signal received from the image color signal converter 301, and calculates a weight by luminance through a luminance weighting algorithm based on the extracted luminance signal, thereby measuring the image quality. The transmission to 400 (S102).

The ROI distance measurer 303 determines a foveation point from the original image signal received from the external memory 500 through a face detection algorithm, and generates a distance from the foviation point to a predetermined point for each image block. To the region of interest weight calculation unit 304.

The region of interest weight calculation unit 304 calculates the weight of the region of interest through the region of interest weighting algorithm to assign the weight differently as the distance from the posi- tion point becomes greater by using the generated distance as a parameter. (S104).

The motion measuring unit 305 receives the original video signal from the external memory 500, compares the received original video signal with a previous video signal previously stored in a memory (not shown), and then calculates a motion vector for each macro block. To the motion weight calculator 306.

The motion weight calculator 306 calculates the weights of the motions of the macroblocks through the motion weighting algorithm based on the motion vectors calculated by the motion measurer 305 and transmits the weights to the image quality measurer 400 (S106). ).

The spatial frequency converter 307 receives the original image signal from the external memory 500, generates a DCT coefficient for each block, and transmits the DCT coefficient to the space-time frequency weight calculator 308.

The space-time frequency weight calculator 308 receives the DCT coefficients of each block generated by the space-frequency converter 307, calculates the weights according to the space-time frequency through the space-time frequency conversion algorithm, and transmits the weight to the image quality measurer 400. It transmits (S108).

The image quality measuring unit 400 receives the weight by the luminance received from the luminance weight calculator 302, the weight by the ROI received from the ROI weight calculator 304, and the motion weight calculator 306. The final weight information is generated by multiplying the weight by the motion and the weight by the space-time frequency received from the space-time frequency weight calculator 308 (S110).

The image quality measuring unit 400 generates final image quality information by multiplying the final weight information by the structural similarity index received from the structural similarity index calculator 200 (S112).

Embodiments of the present invention are not implemented only through the above-described apparatus and / or method, but may be implemented through a program for realizing a function corresponding to the configuration of the embodiments of the present invention, a recording medium on which the program is recorded, and the like. Such implementations may be readily implemented by those skilled in the art from the description of the above-described embodiments.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements of those skilled in the art using the basic concepts of the present invention defined in the following claims are also provided. It belongs to the scope of rights.

According to the above-described configuration, the present invention can be expected to obtain an effect that can achieve a result similar to an evaluation score that is actually directly scored by a person by reflecting human visual characteristics in measuring image quality of multimedia.

According to the present invention, the results are similar to the scores for subjective evaluation of the image quality, and are automatically calculated by the equations. Can be.

The present invention can be expected to use the interface (external memory, etc.) of the terminal to easily implement a full reference (Full Reference) type without additional distortion that may occur in the data acquisition process.

The present invention can be immediately tested in the field using a terminal without complicated equipment using an external memory, and can be expected to be effective as a portable image quality measuring device.

According to the present invention, by providing an original image and a received image using an external memory interface, it is possible to expect an effect that can easily obtain image data while eliminating additional distortion generated during data conversion.

Claims (13)

Receives an original video signal and a received video signal, and uses a structural similarity index indicating a degree of similarity between the original video signal and the received video signal using any one or more information among similarities of luminance, similarity of variance, and degree of covariance. Structural similarity index calculation unit to calculate; A visual weight calculator configured to calculate one or more visual weights by reflecting factors affecting human visual characteristics—information of any one or more of luminance, region of interest, motion, and space-time frequency; And An image quality measuring unit configured to generate final weight information by multiplying the calculated one or more visual weights, and multiplying the calculated structural similarity index by the calculated final weight information to calculate final image quality information Image quality measuring apparatus comprising a. According to claim 1, The structural similarity index is calculated by multiplying the degree of similarity of the luminance, the degree of similarity of the variance, and the degree of the ratio of the covariance, respectively. According to claim 1, The visual weight calculation unit, A luminance weight calculation for receiving the original image signal including a luminance signal and a color signal, extracting the luminance signal from the received original image signal, and calculating a luminance weight using a luminance weighting algorithm based on the extracted luminance signal part; A region of interest weight calculation unit configured to determine a foveation point representing a region of interest for each image block and calculate a region of interest weight through a region of interest weighting algorithm to assign a weight to a distance away from the posi- tion point; A motion weight calculation unit configured to receive the calculated motion vector from a motion measuring unit for calculating a motion vector for each macro block and calculate a motion weight of each macro block through a motion weighting algorithm; And The discrete cosine transform coefficient is received from a spatial frequency transform unit that receives the original image signal and generates discrete cosine transform coefficients for each image block, and receives a space-time frequency weighting factor through a space-time frequency transform algorithm. Space-time frequency weight calculation unit to calculate Image quality measuring apparatus comprising a. The method according to claim 1 or 2, The structural similarity index is

Using X, where X is the pixel of the block in the original image, Y is the pixel of the block in the received image,

Is covariance,

Is the average pixel value of the original image,

Is the average pixel value of the received image,

Is a variance of the original image and the received image, C1, C2 means a constant. The method of claim 3, wherein The luminance weight is

Where Y is the luminance value (L) divided by a predetermined threshold (Lt), and c and r are positive constants that determine the shape of the function. , c = 1, r = 1 / 2.2, threshold value = 50. The method of claim 3, wherein And the ROI weight indicates an extent of resolution drop according to the distance from the ROI. The method of claim 3, wherein And the motion weight is calculated using the magnitude of a motion vector obtained by motion estimation. The method of claim 3, wherein The space-time frequency weights are expressed as a function of time and space frequency.

and

Calculate using

Is the highest spatial and temporal frequency,

Is a normalized discrete cosine transform coefficient, P is a transform exponent for converting the discrete cosine transform coefficients into spatial frequencies, and u and v are coefficients of spatial frequencies. (a) A structural similarity index indicating the degree of similarity between the original image signal and the received image signal by receiving the original image signal and the received image signal from an external memory, and any one of the degree of similarity of luminance, similarity of variance and degree of covariance. Calculate a plurality of visual weight information by using one or more pieces of information and reflecting factors affecting human visual characteristics based on the original image signal—information of any one or more of luminance, region of interest, motion, and space-time frequency. Making; And (b) multiplying the calculated plurality of visual weight information with each other to generate final weight information, and calculating final image quality information by multiplying the calculated structural similarity index with the generated final weight information. Image quality measurement method comprising a. The method of claim 9, In step (a), The original image signal is received from the external memory, and when the received original image signal is an RGB signal, the original image signal is converted into an image signal including a luminance signal and a color signal and output. The received original image signal is the image signal. Outputting as is; And Extracting the luminance signal from the image signal, and calculating a weight by luminance using a luminance weighting algorithm based on the extracted luminance signal; Image quality measurement method comprising a. The method of claim 9, In step (a), Receiving the original video signal from the external memory, a focal point (Foveation Point) is determined through a face detection algorithm from the received original video signal, and generates a distance from the poavation point to a predetermined point for each image block Designating a region of interest; And Calculating weights by the region of interest through the region of interest weighting algorithm to assign weights differently as the distance from the pore point based on the designated region of interest; Image quality measurement method comprising a. The method of claim 9, In step (a), Receiving the original image signal from the external memory, and comparing the received original image signal with a previous image signal previously stored in the memory to calculate a motion vector for each macro block; And Calculating a weight due to the motion of each macro block using a motion weighting algorithm based on the calculated motion vector Image quality measurement method comprising a. The method of claim 9, In step (a), Receiving the original image signal from the external memory and generating discrete cosine transform coefficients for each block; And Calculating weights according to space-time frequency using the discrete cosine transform coefficient of each generated block using a space-time frequency conversion algorithm Image quality measurement method comprising a.

Images