KR20050024098A - Apparatus and Methods for Objective Assessment of Video Quality - Google Patents

Abstract

PURPOSE: An apparatus and a method for estimating video quality using preestimated-value correction are provided to reduce an estimation time, perform real-time estimation and carry out objective estimation according to a predetermined reference. CONSTITUTION: A source video input part(1) receives a source video as an original video. An estimation target video input part(2) receives an estimation target video. A picture quality estimation part(3) estimates picture quality based on the source video and the estimation target video. A correction feature extraction part(4) extracts a feature necessary for correction based on the source video and the estimation target video. An estimation value correction part(5) calculates a correction value by using a correction feature vector which is an output of the correction feature extraction part.

Description

Apparatus and Methods for Objective Assessment of Video Quality}

The present invention relates to an apparatus and method for evaluating video quality using predictive value correction that can objectively evaluate video quality.

In general, the evaluation of the quality of a video is subjective by evaluators who evaluate the quality of the video, and most of these evaluations are performed using the reference video.

In a full reference, the evaluator is called a subjective test of video quality because the evaluator compares the reference video with the video to be evaluated and assigns subjective scores to the video to be evaluated. Lose.

That is, such a conventional evaluation method reflects human perception, and it requires a lot of evaluators to make an evaluation, so that it takes a lot of time, the evaluation cost increases, and the evaluation cannot be performed in real time. .

The present invention is to solve such a problem, an object of the present invention can reduce the evaluation time, as well as to evaluate the image quality of video using predictive value correction that can perform the evaluation in real time and perform an objective evaluation according to a predetermined criterion An apparatus and method are provided.

An apparatus for evaluating video quality using predictive value correction according to the present invention for achieving the above object includes a source video input unit for inputting a source video that is an original video, an evaluation target video input unit for receiving an image to be evaluated, and An image quality evaluation unit for calculating an objective evaluation value (predictive value of subjective image quality evaluation value) of the evaluation target video using the source video inputted to the source image input unit and the evaluation target video inputted to the evaluation target input unit, and inputting to the source image input unit. And a feature extractor for extracting a feature for correction using a source video and an evaluation target video inputted to the evaluation target input unit, and a prediction value correction unit for correcting output data of the image quality evaluation unit.

In the present invention, the prediction value means output data of the image quality evaluation unit, that is, the object quality evaluation value, which corresponds to the prediction value of the subjective image quality evaluation value.

In addition, the video quality evaluation apparatus using the predictive value correction according to the present invention, in the video quality evaluation apparatus, a source image input unit for inputting the source image that is the original video, the evaluation target image input unit for receiving the image to be evaluated An outline extractor for extracting an outline region of the source image input to the source image input unit, a mask image generator for generating a mask image from the output image of the outline extractor, and input into the source image input unit A signal-to-noise ratio calculation unit for calculating an outline maximum signal-to-noise ratio by inputting an image corresponding to the mask image and an image input to the evaluation target input unit among the images to be input; The image quality is evaluated from the output data of the prediction correction unit. Which it is characterized in that it comprises the image quality evaluation unit.

In addition, the image quality evaluation method using the prediction correction according to the present invention, step 1 for extracting the outline in each image (image) of the source video sequence that is the original video, the threshold for generating a mask image from the extracted outline after the outline extraction Setting a value and generating a mask image accordingly, comparing the number of pixels in the outline region of the mask image with a preset reference value and calculating a signal-to-noise ratio when the number of pixels is equal to or greater than the reference value; Step 4, characterized in that it comprises a five step of calculating the evaluation in accordance with the signal to noise ratio.

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

1 is a basic block diagram illustrating a configuration according to the present invention.

Referring to FIG. 1, the video quality evaluation system according to the present invention includes a source video input unit 1 for inputting a source video (moving video), which is an original video, and an evaluation target video for receiving an image to be evaluated. A video quality evaluation unit 3 for performing video quality evaluation using the input unit 2, each of the images input to the source image input unit 1 and the evaluation target image input unit 2, and the source image input unit 1 ) And output data of the video quality evaluation unit by using the correction feature extracting unit 4 for extracting a correction feature from each image input to the evaluation target image input unit 2 and the output data of the correction feature extracting unit 4. And a prediction value corrector 5 for correcting the equation.

The source image input unit 1 and the evaluation target image input unit 2 may be either an input device that receives an image from an external device or a storage device in which an image is stored in advance. The correction feature extractor 4 extracts a correction feature vector, which is one or a combination of energy, blocking degree, edge information, and color information for each region of the space-time frequency, and outputs the correction feature vector to the predictive value corrector 5.

The basic configuration of FIG. 1 may be configured by using a video image quality evaluation unit that measures degradation of a boundary area and evaluates image quality. Hereinafter, an embodiment thereof will be described in detail.

One of the methods for evaluating the quality of a video is to calculate a mean squared error (hereinafter, referred to as 'MSE') between the source video and the video to be evaluated, and accordingly, a peak signal-to-noise ratio. , Hereinafter referred to as PSNR, may be used to evaluate the video.

However, the method of calculating the squared mean error and calculating the maximum signal-to-noise ratio according to the above-described method performs an operation on all the pixels of all the frames of the video, so that the calculation amount is excessive and the load of the apparatus performing the operation is increased. There were disadvantages and inaccurate problems.

On the other hand, the human visual system recognizes that the image quality of the image is good or bad according to the state of the outline of the image included in the image. Therefore, a system and method for measuring the degradation of the image quality near the outline of the moving image and correcting the data (EPSNR) obtained therefrom may be used to evaluate the quality of the moving image. The technique for the objective image quality evaluation apparatus for calculating the EPSNR is described in detail in "Video quality evaluation apparatus and method" (Application No. 10-2003-0013188) filed by the applicant on March 03, 2003.

An embodiment based on the above description is shown in FIG. 2.

FIG. 2 illustrates the embodiment shown in FIG. 1 in more detail. The video quality using the predictive value correcting method according to the present invention includes using a video quality evaluating unit which measures degradation of a boundary area and evaluates image quality. It is a block diagram for demonstrating the structure of an evaluation system.

Referring to FIG. 2, the video quality evaluation system according to the present invention includes a source video input unit 10 for inputting a source video (moving video) which is an original video, and a source video input to the source video input unit 10. A mask image for generating a mask image from the output of the outline extraction unit 30 for extracting the outline of the image, an evaluation target image input unit 20 for receiving an image to be evaluated, and the outline extracting unit 30. A signal-to-noise ratio calculation unit (hereinafter referred to as 'EPSNR') for calculating an outline maximum signal-to-noise ratio (hereinafter referred to as 'EPSNR') of the source image and the evaluation target image using the output of the generator 40 and the mask image generator 40 50), a predictive value correcting unit 60 for correcting the output data of the EPSNR calculating unit 50, and a scaling unit for scaling the output of the predictive value correcting unit 60 at a predetermined ratio ( Including 70) It is. The scaling unit scales the output value according to the user's convenience and has nothing to do with performance.

The source image input unit 10 and the evaluation target image input unit 20 may be an input device that receives an image from an external device or a predetermined storage device in which an image is stored in advance.

In the objective video quality evaluation system according to the present invention, an outline extraction algorithm is applied to a source video to find an outline region, an operation using a threshold is obtained, and an outline region is obtained. By calculating the mean squared error (MSE) of the outline region, the degradation of the outline region, that is, the degree of degradation of the outline is measured. The EPSNR is calculated from the MSE and prediction correction is performed.

In more detail, the outline extractor 30 may use one of the algorithms previously proposed to find the outline region, and may use, for example, a gradient operator. That is, a vertical gradient image is generated by applying an edge extraction algorithm using a vertical gradient operator among the gradient operators to a source image, and at the same time (or after generating), a horizontal gradient image is generated by applying a horizontal gradient operation to the source image. An absolute value gradient image is generated by performing an absolute value operation on the generated vertical and horizontal gradient images.

The mask image generator 40 finds an outline region for MSE calculation by extracting that the pixel value is greater than or equal to a predetermined value from the absolute value gradient image by applying a threshold operation to the absolute value gradient image. That is, only pixels whose values of each pixel of the absolute value gradient image are equal to or larger than a preset threshold value are extracted, and the extracted pixels become an outline region.

Another method of extracting the outline and mask image of an image is as follows. The outline extractor 30 generates a vertical gradient image by applying a vertical gradient operator to the source image, and then generates a continuous vertical horizontal gradient image by applying a horizontal gradient operator to the generated vertical gradient image. Here, it is also possible to generate a horizontal gradient image with respect to the source image and then apply a vertical gradient operator.

As described above, when the vertical horizontal gradient image is generated, the outline extractor 30 transmits it to the mask image generator 40, and the mask image generator 40 applies a threshold value te to the vertical horizontal gradient image. Only pixels having an absolute value greater than or equal to a threshold value among pixels of the gradient image are extracted. In this way, the outline region composed of the extracted pixels becomes a mask image. At this time, the threshold te is initially set to a default value, and then a mask image is generated, and if the threshold number of pixels of the source video outline is not more than a predetermined value in the generated mask image and the threshold value is not less than or equal to a preset lower limit value, The method is repeated while decreasing the predetermined value.

Subsequently, when the number of pixels in the source video outline region is greater than or equal to a predetermined value or the threshold is less than or equal to a predetermined lower limit value, an EPSNR is calculated by comparing the image with the evaluation target image after generating the mask image. To this end, a square mean error (MSE) is used between the outline region of the source image and the target image. That is, the video may be composed of a plurality of frames (or fields), and the plurality of pixels constituting the image of each frame may be represented by a matrix. When the pixels of each frame are represented as a matrix, an error between the source image and the image to be evaluated can be obtained, and the squared mean error (MSE) can be calculated by dividing the obtained error value by the total number of pixels in the outline region, thereby calculating the EPSNR. Will be.

In general, when the boundary region of the evaluation object video is very small compared to the boundary region of the original image, the evaluators tend to perceive the evaluation object video worse. In order to take this into consideration, after calculating the EPSNR, the corrected value MEPSNR may be calculated by performing correction on the EPSNR according to the loss of the boundary area, and the following Equation 1 may be used.

At this time,

EPsrc : the number of pixels of the outline region of the source movie,

EPcommon : The number of pixels in the joint outline area of the source and processed movies.

As described above, if the EPsrc is greater than or equal to a reference value (e.g., for 10,000 pixels, an 8 second video based on NTSC / PAL), the above equation is used; if the EPsrc is less than a reference value of 10,000, the EPsrc is The threshold te is reduced by a predetermined value (for example 20) until it is greater than or equal to 10,000. If the threshold is a lower limit (eg 80), the above correction is not used if the EPsrc is less than 10,000. The correction method based on the loss of the boundary region may be used to correct other objective image quality evaluation methods. In Equation 1, various thresholds may be changed according to an application target.

In addition, in the case of EPSNR, a high value tends to overestimate the subjective evaluation. Therefore, in this case, correct it as follows.

In Equation 2, various thresholds can be changed according to the application target.

In the above embodiment, it is assumed that the image quality evaluation unit which evaluates the image quality by measuring the deterioration of the boundary area is used, but the prediction value correction unit 60 performs the prediction correction even when an arbitrary image quality evaluation unit is used. .

The predictive value correcting unit 60 according to the present invention corrects the output data of the EPSNR calculation unit 50, that is, the image quality evaluation unit. The predictive value correction method is a process in which the objective image quality evaluation system can process the prediction more accurately. In the video objective image quality evaluation system, predictive value correction is a method of correcting a predicted value (objective evaluation value) so as to approximate the actual value from the actual value (subjective evaluation value) for an arbitrary prediction method.

Prediction correction method according to the present invention for extracting the correction features (correction vectors) to find the optimal linear weight vector (Least Mean Square (LMS) method, rule-based method, neural network, etc.) Can be used. The least mean square method described above Find the weight vector a so that is minimized. Where Y is the correction feature vector used for the correction, b is the difference between the actual value and the predicted value, and a is the weight vector. The weighted vector is multiplied by the correction feature vector to calculate the correction value and adds it to the predicted value so that it is close to the actual value.

The features that can be used for the correction include the difference in space-time frequency for each region, blocking (blcoking) measurement, edge (boundary line) information, color difference, etc., and other features can be set and used. In particular, the wavelet coefficient may be used to obtain the difference of the space-time frequency for each region.

The wavelet coefficients may directly use wavelet coefficients as a result of wavelet transforming an image, or may use energy of each frequency domain or change information according to time of energy.

Blocking phenomenon occurs in block-based image encoding using Discrete Cosine Transform. The degree of blocking is one of the factors affecting the image quality. Can be.

The boundary region of the image is particularly sensitive to the human eye, and the distortion degree (edge damage) of the boundary region can be used as a correction feature vector.

In general, a method of using color difference is to use information of three colors of RGB constituting an image as a correction feature vector.

A correction feature vector is constructed using the above features. For example, the correction feature vector y ₀ can be constructed using the energy of each region of the wavelet coefficient, and the energy, the degree of blocking (y ₁ ), the edge information (y ₂ ), the color information (y ₃ ), etc. Can be combined to form a feature vector.

In other words, each feature y ₀ , y ₁ , y ₂ , y ₃ ... is composed of the correction feature vector y and multiplied by the appropriate weight vector a to calculate the correction value b as shown in Equation 3 below. .

a ^T y = b

In this case, the image quality evaluation and correction should be optimized for a large number of videos, not just one video, so that a matrix can be constructed using multiple feature vectors. That is, in the case of using n moving images, the matrix Y composed of the feature vectors is expressed by Equation 4.

Accordingly, as described above Calculate the value of the weight vector a such that is minimized. Where b is the difference vector representing the difference between the actual value and the predicted value. In addition, a method of estimating a correction value as a nonlinear function of a correction feature vector may be used. For example, a polynomial or a neural network may be used.

In the case of using the linear weighting value, when the weighting vector is calculated, the calculated weighting vector is applied to the data calculated by the image quality evaluation unit by applying the calculated weighting vector to the correction feature vector, thereby performing predictive value correction.

3 is a graph illustrating a prediction value correction method according to the present invention.

As shown in FIG. 3, the above-described predictive value correction method is to correct data such that values that are separated from the actual value among the data are close to the actual value.

As shown in Fig. 3, it can be seen that the predicted value after the predictive value correction is closer to the actual value than before the predicted value correction.

On the other hand, the prediction value correction method described above may divide the output of the image quality evaluation unit into predetermined sections as shown in FIG. 4 and apply different correction methods to each section.

That is, a plurality of windows (three window0, window1, and window2 in this embodiment) are set according to the value of the output data of the image quality evaluation unit. In addition, a correction method may be applied to each window.

The method of the present invention is described below using a flowchart on the assumption that an image quality evaluation unit for evaluating image quality by measuring deterioration of a boundary area is as follows. In the following embodiment, the boundary as shown in FIG. Although the configuration using the image quality evaluation unit for measuring the deterioration of the area and evaluating the image quality is used, it will of course be possible to use any image quality evaluation method according to the basic configuration of FIG.

5 is a flowchart illustrating an embodiment of a video objective image quality evaluation method using a predictive value correction method according to the present invention on the assumption that an image quality evaluation unit is used to measure image quality by measuring degradation of a boundary area.

Referring to FIG. 5, first, an outline is extracted from a source image (S10). To this end, the outline extraction unit 30 extracts the outline of the source image input through the source image input unit 10 shown in FIG. 2. For outline extraction, for example, an outline extraction algorithm using a gradient operator may be applied. That is, a vertical gradient operator and a horizontal gradient operator are applied to the source image to generate a vertical gradient image and a horizontal gradient image, and an absolute value operation is performed to extract the outline, or to apply a vertical gradient operator to the source image to apply a vertical gradient. By generating an image and applying a horizontal gradient operator to a vertical gradient image to generate a continuous horizontal gradient image, the absolute value calculation can be performed to extract the outline of the source image.

A threshold te for generating a mask image is set to a predetermined value, for example te = 260 in this embodiment (S20). After setting the threshold value, a threshold value is applied to the outline image and a mask image is generated (S30). The mask image is generated by the mask image generator 40. To generate a mask image, threshold values are applied to the vertical and horizontal gradient images generated by the outline extracting unit, and the absolute value of the pixels forming the outline is greater than or equal to a predetermined value and output as a mask image.

In operation S40, it is determined whether the number of outline area pixels EPsrc of the source video after the mask image is greater than or equal to the reference value (for example, 10000) or the threshold value is less than the lower limit (for example, 80) (S40).

In operation S40, if the number of pixels of the outline region of the source video is not greater than or equal to the reference value, and the threshold value te is less than the preset lower limit value, the threshold value is reset (S41). At this time, the threshold value is reset by decreasing the predetermined value from the previous value.

If the condition is satisfied in the above step, the EPSNR calculation unit 50 calculates the EPSNR using the mask image and the evaluation target image (S50). That is, an error between the source image of the outline region and the evaluation target image may be obtained using the mask image of the source image. The error values obtained in this way are divided by the total number of pixels in the outline region to calculate a squared mean error (MSE), and the EPSNR can be calculated therefrom.

After the EPSNR is calculated, the correction is performed. The correction value MEPSNR is calculated using the following equation according to the above-described method.

At this time,

EPsrc : the number of pixels of the outline region of the source movie,

EPcommon : The number of pixels in the joint outline area of the source and processed movies.

Accordingly, the prediction value correcting unit 60 may perform prediction correction by the above-described method (S60). In operation S70, the scaling unit 70 performs scaling on the calculated data according to the prediction value correction. The scaling unit 70 scales the output value according to the user's convenience and has nothing to do with performance.

The prediction value correction process of step S60 will be described with reference to FIG. 6. 6 is a flowchart illustrating a prediction value correction method according to the present invention.

As described above, the predictive value corrector 60 according to the present invention first constructs a correction feature vector for predictive correction (S71). For example, a correction feature vector (y ₀ ) can be constructed using energy of each region of space-time frequency, and such energy, blocking degree (y ₁ ), edge information (y ₂ ), color information (y ₃ ), etc. Can be combined to construct a correction feature vector.

That is, a correction feature vector y is composed of each feature y ₀ , y ₁ , y ₂ , y ₃ ..., and the like, and a correction function to be used is calculated (S72). A linear function can be used as the simplest correction function, and in this method, the predicted value b is calculated by multiplying an appropriate weight vector a such that a ^T y = b (S73). Other correction functions can be used for nonlinear functions such as neural networks, polynomials, and so on.

7 is a graph illustrating the correlation between the results of the objective video quality evaluation method and the subjective video quality evaluation method according to the present invention.

Referring to FIG. 7, (a) shows a correlation according to a method of calculating EPSNR, and (b) shows a correlation after correlation according to the present invention. Looking at the graph (b) of the result according to the method according to the present invention, it can be seen that the distance between each data is closer and the correlation is higher than the data of the result according to the general PSNR calculation method. have.

As described above, according to the present invention, the correlation of the data is higher than that of the conventional method, and a more accurate subjective evaluation value can be predicted by the method of correcting the predicted value.

As described in detail above, according to the apparatus and method for evaluating moving image quality using prediction value correction according to the present invention, it is possible to predict subjective evaluation values more accurately and to perform objective evaluation according to predetermined criteria in real time.

1 is a basic block diagram for explaining the configuration of a video quality evaluation apparatus using predictive value correction according to the present invention;

FIG. 2 is a block diagram for explaining a configuration of applying a moving image quality evaluation unit for measuring image quality by measuring degradation of a boundary region in FIG. 1;

3 is a graph for explaining a predictive value correction method according to the present invention;

4 is a graph for explaining another embodiment of a predictive value correction method according to the present invention;

5 is a flowchart illustrating a video quality evaluation method using predictive value correction according to the present invention;

FIG. 6 is a flowchart for explaining the prediction value correcting step of FIG. 5 in detail;

7 is a graph illustrating a result according to a video quality evaluation method using predictive value correction according to the present invention;

<Description of Symbols for Main Parts of Drawings>

10: source image input unit 20: evaluation target image input unit

30: outline extraction unit 40: mask image generation unit

50: EPSNR calculation part 60: Prediction

70: scaling part

Claims (27)

In the video quality evaluation apparatus, A source video input unit for inputting a source video, which is an original video, Evaluation target image input unit for receiving the image to be evaluated, An image quality evaluation unit for evaluating image quality by using the source image inputted to the source image input unit and the evaluation target image inputted to the evaluation target image input unit;  A correction feature extracting unit for extracting a feature required for correction by using the source image inputted to the source image input unit and the evaluation target image inputted to the evaluation target image input unit; An apparatus for evaluating moving picture quality using predictive value correction, characterized in that a predictive value correction unit calculates a correction value using a corrected feature vector output from the corrected feature extractor. The method of claim 1, And the source image input unit and the evaluation target image input unit are any one of an input device for receiving an image from an external device or a storage device in which an image is stored in advance. The method of claim 1, And a correction feature vector, which is an output of the correction feature extracting unit, is any one of energy, blocking degree, edge information, and color information of each space-time frequency, or a combination thereof. In the video quality evaluation apparatus, A source video input unit for inputting a source video, which is an original video, Evaluation target image input unit for receiving the image to be evaluated, An outline extracting unit for extracting an outline region of the source image inputted to the source image input unit;  A mask image generator for generating a mask image from the output image of the outline extractor; A signal-to-noise ratio calculation unit configured to calculate an outline maximum signal-to-noise ratio by inputting an image corresponding to the mask image and an image input to the evaluation target input unit among the images input to the source image input unit; And a predictive value corrector for correcting the output data of the signal-to-noise ratio corrector. The method of claim 4, wherein And the source image input unit and the evaluation target image input unit are any one of an input device for receiving an image from an external device or a storage device in which an image is stored in advance. The method of claim 4, wherein And a scaling unit configured to scale and output the output of the predictive value corrector at a predetermined ratio. The method of claim 4, wherein The outline extracting unit extracts an outline region by applying an outline extraction algorithm by a gradient operation to an image of an input video and transfers the image to the signal-to-noise ratio calculating unit. According to claim 4, The mask image generating unit sets a pixel whose absolute value of the pixel of the outline extracting unit output image is less than or equal to a preset threshold value to zero, and sets a pixel greater than or equal to the threshold value to a non-zero value. . The method of claim 8, The mask image generation unit sets a threshold value to a predetermined value, generates a mask image by applying the threshold value, and if the number of pixels in the outline area of the mask image is less than or equal to a preset reference value, the mask image generator uses the current mask image as the final mask image. Apparatus for evaluating video quality using predictive value correction, characterized in that output to the noise ratio generator. The method of claim 9, The mask image generator may reduce the threshold by a predetermined value and regenerate the mask image when the number of pixels of the outline area of the mask image is not greater than or equal to a preset reference value and the threshold is greater than or equal to a preset lower limit. Apparatus for evaluating video quality using predictive value correction. The method of claim 10, And the mask image generator outputs the current mask image as a final mask image to the signal-to-noise ratio generator if the threshold value is less than or equal to a preset lower limit value. According to claim 7, The signal-to-noise ratio calculation unit calculates a difference between a value between an outline region of the source video and each pixel of the same sequence of the evaluation target video and calculates an average by dividing the total number of pixels of the outline region of the outline video sequence by the total value. Apparatus for evaluating video quality using predictive value correction, characterized in that the. The method of claim 4, wherein And the predictive value correcting unit calculates a correction value by comparing the boundary region area of the source image with the boundary region area of the evaluation target image. The method of claim 4, wherein The prediction value compensator configures a correction feature vector with respect to the output data of the signal-to-noise ratio correction section, and performs correction by calculating a correction value using the correction feature vector and a predetermined correction function to perform prediction correction. Apparatus for evaluating video quality using predictive value correction. The method of claim 14, The apparatus of claim 1, wherein the correction feature vector is any one of energy, blocking degree, edge information, and color information of each region of space-time frequency, or a combination thereof. The method of claim 14, The predictive value correction section divides the output data of the signal-to-noise ratio correction section into predetermined sections, constructs a correction feature vector, and performs correction by applying different correction functions to each section. Image quality evaluation device. In the video quality evaluation method using prediction correction, Step 1 extracts the outlines of each image in the source video sequence that is the original video, Setting a threshold value for generating a mask image from the extracted outline after extracting the outline, and generating a mask image accordingly; Comparing the number of pixels in the outline region of the mask image with a preset reference value and calculating a signal-to-noise ratio when the number of pixels is greater than or equal to the reference value; And a four step of correcting the signal-to-noise ratio. The method of claim 17, The algorithm for extracting the outline is a video quality evaluation method using prediction correction, characterized in that using a gradient operator. The method of claim 17, The signal-to-noise ratio is calculated using a squared mean error, The calculation of the squared mean error may include calculating a difference between pixels of the outline regions between the image of the source video sequence and the evaluation target video sequence image, and adding the results of the first step to the entire frame to determine the area of the outline region. And calculating the total sum of squared errors and dividing the total sum by the number of pixels in an outline region. The method of claim 19, The squared mean error is a video using predictive value correction, wherein the pixel value of the mask image of all the pixels in the outline regions between the source video sequence and the evaluation target video sequence is equal to or greater than the preset threshold value. Image quality evaluation method. The method of claim 17, In step 3, comparing the number of pixels in the outline region of the mask image with a preset reference value, if the number of pixels is not equal to or greater than the reference value, the threshold value is decreased by a predetermined value, and in step 2, the mask image is regenerated. Method for evaluating video quality using predictive value correction, characterized in that the. The method of claim 21, In the third step, when the threshold value is less than a predetermined lower limit value, the threshold value is set to the lower limit value, and the signal-to-noise ratio is calculated. The method of claim 17, Step 4 is a video quality evaluation method using the prediction correction, characterized in that to calculate the correction value by comparing the boundary area of the source image and the area of the evaluation target image. The method of claim 17, The fourth step, Constructing a correction feature vector with respect to the result of step 3; Calculating a correction function for the correction feature vector and applying the correction function to the correction feature vector to calculate a correction value; And a step of correcting the result of the three steps using the correction value. The method of claim 24, The correction feature vector is any one of a combination of energy, blocking degree, edge information, color information for each region of space-time frequency, or a combination thereof. The method of claim 24 In calculating the correction feature vector in step 4, the result of step 3 is divided into predetermined sections, and different correction values are calculated for each section to calculate different correction values for each section. How to evaluate video quality. The method of claim 17, The image quality evaluation method further comprises the step of scaling the result of the step 4 by a predetermined ratio.