KR20110041962A - Method and apparatus for video quality assessment - Google Patents

Abstract

PURPOSE: A method and device thereof are provided to measure image quality without any information about an original image by using information included in mage data within a bit string inside a receiving side. CONSTITUTION: An image quality measuring method includes as follow: a step of receiving video data; a step of obtaining a discrete cosine transform coefficient and a quantization parameter included in video data; a step of generating a frequency component histogram of video data by using the discrete cosine transform coefficient; a step of generating a histogram which models to a Gaussian function histogram; a step of computing a quantization distortion amount per frequency component of video data by using the modeled histogram modeled and the quantization parameter.

Description

METHOD AND APPARATUS FOR MEASUREMENT OF IMAGE QUALITY {METHOD AND APPARATUS FOR VIDEO QUALITY ASSESSMENT}

The present invention relates to a method and apparatus for measuring image quality, and more particularly, to a method and apparatus for measuring image quality using only information included in a bitstream without an original image.

Recently, with the development of wired and wireless communication technology, digital video services such as VoD (Video on Demand) and IPTV (Internet Protocol TeleVision) are being provided. However, because the bandwidth of digital moving image data is too large, in order to provide a service through a communication network, the bandwidth of the video data must be compressed, and the image quality is often lost during the bandwidth compression of the video. Therefore, in order to provide a higher quality image service, it is necessary to measure the quality of the provided image.

Image quality measurement methods are classified into full reference, reduced reference, and no reference methods. Electron Jun is a method of measuring the image quality using the original image, and the reduction criterion method is a method of measuring the image quality using only the characteristics of the original image, and the non-standard method measures the image quality without information about the original image. That's how.

The conventional non-standard video quality measurement method mainly measures the blockiness, blur, jerkiness, etc. of the decoded video. A nonstandard video quality measurement method is to find and score a feature of a degraded image from the decoded image without information on the original image. Blockiness is a score of the degree of boundary generation in 8x8 block units, one of the characteristics of deteriorated image. Blur scores the degree of crushing of edges, one of the characteristics of degraded images. Jerkiness scores the degree of frame rate drop among the characteristics of degraded images.

However, these blockiness, blur, and jerkiness measurements often do not accurately reflect the actual image quality. For example, in the case of Blur measurement, the difference between the original borderless image and the borderless image due to deterioration of image quality cannot be distinguished. In the case of Jerkiness, the frame rate is lower than the original borderless motion. It is difficult to tell the difference between no motion. There are other non-standard video quality measurement methods, but the performance is poor because it basically measures the quality without the original video.

An object of the present invention is to provide an image quality measuring method and apparatus capable of measuring the quality of an image without information on an original image by using information included in image data in a bit string at a receiver.

The present invention also provides an image quality measuring method and apparatus capable of measuring image quality more accurately than the conventional non-standard method by measuring the image quality using information included in the image data in the bit stream before decoding the image data. For other purposes.

Another object of the present invention is to provide a method and apparatus for measuring image quality, which is capable of measuring image quality more accurately by reflecting visual characteristics of a viewer watching a corresponding image by assigning different weights to each region when measuring distortion of an image. It is done.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention, which are not mentioned above, can be understood by the following description, and more clearly by the embodiments of the present invention. Also, it will be readily appreciated that the objects and advantages of the present invention may be realized by the means and combinations thereof indicated in the claims.

In order to achieve the above object, the present invention provides a method for measuring image quality, comprising: receiving image data, obtaining discrete cosine transform coefficients and quantization parameters included in the image data, and using the discrete cosine transform coefficients. Generating a histogram for each frequency component, generating a histogram that models the generated histogram with a Gaussian function, and calculating a quantization distortion amount for each frequency component of the image data using the modeled histogram and quantization parameters. Characterized in that it comprises a.

In addition, the present invention provides an image quality measuring apparatus comprising: an input unit for receiving image data, a data management unit for obtaining discrete cosine transform coefficients and quantization parameters included in the image data, and a histogram for each frequency component of the image data using discrete cosine transform coefficients And a histogram modeling the generated histogram with a Gaussian function, and using the modeled histogram and quantization parameters, an operation unit for calculating the quantization distortion amount for each frequency component of the image data. do.

The present invention also provides a method for measuring image quality, comprising: receiving image data divided into a central region and a peripheral region, calculating a quantization distortion amount for the central region and a peripheral region, a quantization distortion amount for the central region, and a peripheral region Weighting the amount of quantization distortion for the region, and summing the amount of quantization distortion for each frequency component for the weighted center region and the amount of quantization distortion for each frequency component for the peripheral region. It is done.

According to the present invention as described above, by using the information contained in the image data in the bit stream at the receiver side there is an advantage that can measure the quality of the image without information on the original image.

In addition, according to the present invention, by measuring the quality of the image by using the information contained in the image data in the bit stream before decoding the image data, there is an advantage that the accurate image quality can be measured than the conventional non-standard method.

In addition, the present invention has the advantage that it is possible to measure the image quality more accurately reflecting the visual characteristics of the viewer watching the image by assigning different weights to each region when measuring the distortion of the image.

The above objects, features, and advantages will be described in detail with reference to the accompanying drawings, whereby those skilled in the art to which the present invention pertains may easily implement the technical idea of the present invention. In describing the present invention, when it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description will be omitted. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to indicate the same or similar components.

1 is a configuration diagram of a communication network for explaining the operation of the image quality measuring apparatus according to the present invention.

Referring to FIG. 1, a video signal (original video) provided by a broadcaster, a broadcast content provider, or the like is encoded by the encoder 102. The encoded image data (hereinafter, referred to as image data) output in the form of a bit string by the encoder 102 is transmitted to the receiving side through the communication network 104. Here, the image data mainly has the form of a bit stream. The decoder 106 on the receiving side receives and decodes the transmitted video data, and the decoded video data is transmitted to the playback device 108 for reproduction.

In this case, the methods of measuring non-standard video quality by measuring Blockiness, Blur, Jerkiness, etc. as described above measure the quality of the video using the decoded video, that is, the decoded video data output from the decoder 106. . In this case, a problem as described above, that is, a problem of not accurately measuring the quality of the actual image occurs.

In order to solve this problem, the image quality measuring method according to the present invention can obtain a more accurate quality measurement result by measuring the quality of the image using the image data before decoding. In FIG. 1, the image quality measuring apparatus 110 according to the present invention receives image data transmitted through the communication network 104, obtains useful information included in the image data, and uses it for image quality measurement.

Here, useful information that the image quality measuring apparatus 110 can obtain through the image data includes various header information, discrete cosine transform (DCT) coefficients, quantization parameters, the number of bits actually used, and the like. Can be mentioned.

The image quality measuring apparatus 110 more accurately measures and outputs the quality of the transmitted image by using the information. The output measurement result may be transmitted to the transmitting side, for example, the encoder 102 as shown in FIG. 1, thereby helping the transmitting side to perform a better image providing. Although not shown in FIG. 1, the image quality measurement result of the image quality measuring apparatus 110 may be provided to another device such as a decoder 106. In addition, the decoded image output by the decoder 106 may be provided to the image quality measuring apparatus 110 and used for various purposes such as comparing an actual image with a measurement result.

2 is a configuration diagram showing the configuration of an image quality measuring apparatus according to the present invention.

Referring to FIG. 2, the image quality measuring apparatus 202 includes an input unit 204, a data manager 206, and a calculator 208.

In an embodiment of the present invention, the input unit 204 receives image data transmitted through a communication network or other medium. The data manager 206 acquires information included in the image data input through the input unit 204, for example, a discrete cosine transform coefficient, a quantization parameter, and the like.

On the other hand, the calculator 208 calculates the amount of distortion of the image data by using the information obtained by the data manager 206 and other information. Specifically, the operation unit 208 generates a histogram for each frequency component of the image data by using the discrete cosine transform coefficients, generates a histogram modeling the generated histogram with a Gaussian function, and generates the modeled histogram and the quantization parameter. The amount of quantization distortion for each frequency component of the image data can be calculated.

In order to calculate the amount of quantization distortion for each frequency component, the calculating unit 208 divides the quantization step sections set in the modeled histogram into n detailed sections, where n is a positive integer, and the detailed section of the quantization step sections. A function value for each quantization step may be calculated by calculating a function value for each quantized step, using the calculated function value for each subdivision section, and adding the distortion amount of the quantization step section.

In another embodiment of the present invention, the input unit 204 receives image data divided into a central area and a peripheral area. The data manager 206 acquires information included in the image data input through the input unit 204, for example, a discrete cosine transform coefficient, a quantization parameter, and the like.

The calculator 208 calculates the amount of distortion of the image data by using the information obtained by the data manager 206 and other information. Specifically, the calculating unit 208 calculates the quantization distortion amount for the center region and the peripheral region, weights the quantization distortion amount for the central region and the quantization distortion amount for the peripheral region, respectively, and the weighted center region. By summing the quantization distortion amount for each frequency component with respect to the quantization distortion amount for each frequency component with respect to the surrounding area, a more accurate image quality measurement result reflecting human visual characteristics can be obtained. Herein, the operation of calculating the quantization distortion for the center region and the peripheral region by the calculator 208 may proceed similarly to the above-described embodiment of the present invention.

Hereinafter, the image quality measuring method according to the present invention will be described in more detail with reference to specific embodiments.

(1) Calculation of distortion by quantization

In the image quality measuring method according to an exemplary embodiment of the present invention, the quality of an image is measured based on no reference by using information obtained from the image data and characteristics of the image. As described above, the receiver may acquire various information such as various header information, DCT coefficients, motion vectors, etc. from the video data before decoding the video data transmitted through the communication network. Image pixels are made by decoding the image data. In this case, the image pixels generated by decoding the image data include distortion, because the distortion includes distortion generated by quantization in the DCT coefficients included in the image data. Therefore, under the assumption that there is no error in the transmission process, distortion due to quantization of the DCT coefficients may be a cause of image quality degradation. Accordingly, if the amount of quantization distortion included in the DCT coefficient can be accurately estimated, the quality of the corresponding image can be more accurately calculated without the original image.

First, image data is input (S302). In operation S304, DCT coefficients and quantization parameters included in the image data are obtained. The DCT coefficients included in the image data are divided into DCT coefficients of intra mode and DCT coefficients of prediction mode. The DCT coefficient of the intra mode is a value obtained by performing discrete cosine transform on the pixel value of the original image, and the DCT coefficient of the prediction mode is a value obtained by performing discrete cosine transform on the difference between the reference image pixel value and the current image pixel value.

Next, a histogram for each frequency component of the image data is generated using the DCT coefficients (S306). FIG. 4 is a histogram for each frequency component generated after dividing a prediction mode image into 4 × 4 blocks and performing discrete cosine transform and quantization of DCT coefficients accordingly. Encoder conforming to MPEG, H.264, etc. divides the video into 8 × 8 or 4 × 4 sized blocks and encodes two-dimensional discrete cosine. In FIG. 4, the upper left component corresponds to a DC component (frequency (0,0) position) of a 4 × 4 block, which is a histogram obtained by collecting only (0,0) frequency components from 21600 4 × 4 blocks. In FIG. 4, the frequency in the horizontal direction increases as goes to the right as (0,1), (0,2), and (0,3), and toward the bottom, as in (1,0), (2,0), (3, As in 0), the frequency in the vertical direction increases.

When the difference between the reference picture pixel value and the current picture pixel value is obtained in the prediction mode, a difference between images having similar pixel values is obtained. As shown in FIG. 4, when the histogram for the DCT coefficients is obtained in the prediction mode, the value of 0 is the most. When the value of 0 is large, it is easy to compress the image. When encoding the image, the encoder encodes most of the image in the prediction mode rather than the intra mode in order to increase the compression performance.

 As shown in FIG. 4, when the histograms of all the frequency components of the prediction mode are drawn, it can be seen that the frequency near zero is the highest, and the frequency decreases as the distance is far from zero.

FIG. 5 is a histogram for each frequency component generated after dividing an intra mode image into 4 × 4 blocks and performing discrete cosine transform and quantization of DCT coefficients accordingly.

Referring to FIG. 5, since the DC component of the intra mode is an average value of pixels, the histogram of the DC component has the highest frequency at the non-zero average pixel value of the image, and decreases as the distance increases from the average pixel value. The histogram of the remaining frequency components has the highest frequency at the center and the frequency decreases toward the periphery.

Next, a histogram in which the generated histogram is modeled as a Gaussian function is generated (S308). It can be seen that the histogram for each frequency component shown in FIGS. 4 and 5 has a form very similar to a Gaussian distribution. Therefore, in the present invention, the histogram for each frequency component of the intra or prediction mode obtained as described above is modeled as a Gaussian function. 6 shows a result of modeling a histogram of frequency components of DCT coefficients using a Gaussian function.

 A Gaussian function for modeling the histogram for each frequency component of the DCT coefficient is shown in [Equation 1].

는 각 주파수 성분별 히스토그램의 표준편차 값이다. 구 체적으로, 예측 모드에서는 참조 영상과 현재 영상간의 화소의 차(residue)의 각 주파수 성분별 히스토그램에 대해 계산한 표준편차이며, 인트라 모드에서는 원래 영상의 화소의 각 주파수 성분별 히스토그램에 대해 구한 표준편차이다.In [Equation 1],

Is the standard deviation value of the histogram for each frequency component. Specifically, in the prediction mode, it is a standard deviation calculated for the histogram for each frequency component of the pixel difference between the reference image and the current image. Deviation.

[Equation 2] is an equation for obtaining the k value of [Equation 1].

In Equation 2, max (H (i)) is the maximum frequency of the histogram of the quantized DCT coefficients. Where k is a factor that determines the height of the middle part when the Gaussian function of Equation 1 is graphed. The k value of Equation 2 is an ideal value obtained through experiments.

Finally, the amount of quantization distortion for each frequency component of the image data is calculated using the modeled histogram and the quantization parameter (S310). 4 and 5 described above are histograms of DCT coefficients before quantization is performed. However, since the decoder can only obtain DCT coefficients after quantization, the distortion amount due to quantization can be accurately calculated only by predicting the histogram before quantization. To this end, a histogram modeled using Equation 1 is used.

7 shows a comparison of a histogram before quantization and a histogram after quantization of DCT coefficients.

In FIG. 7, the histogram represented by a dot represents a histogram after quantization, and the histogram composed of a curve represents a histogram of DCT coefficients before quantization, modeled through Equation 1. 7 is a DCT coefficient value, and a vertical axis is the frequency of DCT coefficients. Since DCT coefficient values are mapped to the nearest discrete quantization values when quantized, DCT coefficient values before quantization appear continuously but discontinuously after quantization.

In an embodiment of the present invention to be described with reference to FIG. 7, 15 quantization step values are provided, and black dots in FIG. 7 indicate the frequency after quantization. The reason why the frequency after quantization is larger than the frequency of continuous values before quantization is that the sum of the total powers of the DCT coefficients must be the same as before and after quantization. That is, since the continuous values between the quantization step values are mapped to both quantization step values, the frequency of the quantization step values becomes larger than the continuous value before quantization.

FIG. 8 is an enlarged view of a portion of FIG. 7 to explain a method of calculating quantization distortion.

8 shows quantization step values (i-1) Qstep, iQstep, and (i + 1) Qstep, and the steps from (i-1) Qstep, iQstep to iQstep and (i + 1) Qstep are Qstep, That is, it is defined as a quantization step interval.

In FIG. 8, DCT coefficient values located between the sections indicated by Qstep are mapped to iQstep values in the center of the Qstep section by quantization, so that the number of times after quantization corresponding to the iQstep value increases by Δn. In other words, the Δn DCT coefficients were not originally iQstep values but were mapped to iQstep values by quantization. Accordingly, the Δn DCT coefficients include quantization distortion by the difference between the original value and the iQstep value. Accordingly, in order to calculate the distortion amount of the Qstep, that is, the quantization step interval, the distortion amount of the Δn DCT coefficients must be calculated.

Since the Δn DCT coefficients can be assumed to be evenly distributed in the quantization step interval, the quantization step interval is divided into n subdivision intervals (where n is an integer) to obtain a distortion amount for each subdivision interval, and the sum is added to the quantization step. Find the amount of distortion for the interval.

9 is a diagram for describing a method of obtaining a Gaussian function value by dividing a quantization step interval into detailed intervals.

In an embodiment of the present invention, as illustrated in FIG. 9, the quantization step interval Qstep is divided into eight subdivisions. Therefore, the interval of the detailed section will be Qstep / 8 as shown in FIG. 9. In another embodiment of the present invention, the accuracy of calculation can be improved by dividing the subdivided interval into smaller sections.

In FIG. 9, the Δn DCT coefficients are distributed along the Gaussian function h (x) of [Equation 1] rather than uniformly distributed in each subdivision section. Therefore, a Gaussian function value should be obtained for each subdivision interval. In this case, x value as shown in [Equation 3] is substituted into h (x) defined in [Equation 1] according to the detailed section.

In Equation 3, j has a value of 1 to 8, that is, the order of subdivisions. In an embodiment of the present invention, since the quantization step interval is divided into eight equal parts, n = 8. [Equation 3] can be obtained x value in each sub-section, and by substituting this x value in [Equation 1] to obtain the h (x) value in each sub-section.

FIG. 10 is a diagram for describing a method of calculating a distortion amount of a quantization step section by calculating a distortion amount of each subdivision section.

As described through Equation 3, a function value in each sub-section j is obtained, and the function value is represented by h ₁ , h ₂ ,... , h ₈ , the distortion amount in each detailed section can be obtained. As shown in FIG. 10, h ₁ and h ₈ are Qstep / 2 (= 4Qstep / 8), h ₂ and h ₇ are 3Qstep / 8, h ₃ and h ₆ are Qstep / 4 (= 2Qstep / 8), h ₄ and h ₅ have a distortion amount of Qstep / 8.

As a result, the distortion amount of the quantization step section Qstep can be obtained by calculating the function values h ₁ to h ₈ for each of the eight detailed sections. The equation for calculating the distortion amount of the quantization step section is shown in [Equation 4].

In Equation 4, k denotes a subscript indicating n subdivisions, d _k represents a distortion amount of the kth subdivision, and D _f represents a distortion amount of one quantization step interval. When the meaning of Equation 4 is explained, Δn DCT coefficients are distributed for each subdivision section by the ratio of the Gaussian function h _k value, and then the amount of distortion for each subdivision section is obtained.

Accordingly, the amount of distortion for one frequency component may be added to each of the quantization step intervals by calculating the distortion as shown in [Equation 4]. In addition, the distortion amount of the entire image can be obtained by adding all the distortion amounts for each frequency component.

(2) Weighted distortion calculation that reflects human visual characteristics

When a person is watching a TV or monitor screen, the focus is mainly on the center, and relatively few corners of the screen are relatively small. If you look at the corner of the screen, you want to know if the glass on your TV or monitor is broken, not because of the screen, or if a specific object is located in the corner of the screen. Unless you're looking for a hidden picture, however, the videographer angles the camera so that the TV viewer sees the center of the screen, and when the object of interest moves to the corner, the camera moves back along the object, bringing the object of interest back to the center of the screen. do.

Accordingly, the viewer's eyes perceive the corner of the screen as the background and do not focus on it in detail. However, since the viewer cannot know where to focus and see in detail, the human vision is sensitive to a certain area of the center, and if a distortion occurs in this area, the distortion is higher than the corner distortion. It needs to be weighted.

11 illustrates distortions of various images in order to experiment with how the human eye perceives differently according to the position of the screen even when the same amount of distortion occurs. The left images 1102 and 1106 distorted the center, and the right pictures 1104 and 1108 distorted the periphery of the image. Through these experiments, it can be seen that even if the same amount of distortion, the viewer perceives the distortion more when the distortion is applied to the center region than to the distortion of the peripheral region.

12 is divided into a central region and a peripheral region in order to give a weight reflecting human visual characteristics.

In an embodiment of the present invention, as described above, the screen is divided into a central region and a peripheral region to reflect human visual characteristics that are more sensitive to the center portion than the corners of the screen, and the distortion amount is applied to each of the central region and the peripheral region. The calculations are performed to give weights to the calculated distortion amount of the center region and the distortion amount of the peripheral region, respectively. At this time, the weight given to the distortion amount in the center region is set to be relatively larger than the weighting value applied to the distortion amount in the peripheral region, thereby reflecting the visual characteristics of the human being more responsive to the distortion in the center portion of the screen. Can be.

First, image data divided into a central area and a peripheral area is received (S1202). There may be various methods for dividing the screen into a central area and a peripheral area. However, in one embodiment of the present invention, as shown in FIG. 10, the screen is divided into six equal parts and the vertical size of the screen is divided into five equal parts. Divide into blocks In this case, the 12 blocks in the middle may be defined as the central region, and the remaining 18 blocks may be defined as the peripheral region. It should be noted that the definition of the screen division, the central region, and the peripheral region may vary depending on the embodiment.

Thus, the distortion amount calculation described in section (1) is performed on the image data divided into the center area and the peripheral area (S1204). At this time, the distortion amount is calculated separately for the 12 blocks in the center region and 18 blocks in the peripheral region, and weighted for each of the calculated distortion amount in the central region and the distortion amount in the peripheral region (S1206). Then, the quantization distortion amount for each frequency component of the weighted center region and the quantization distortion amount for each frequency component of the peripheral region are added (S1208). Since the method of obtaining the distortion amount of the center region and the distortion amount of the peripheral region has been described through Section (1), detailed description thereof will be omitted.

[Equation 5] is a formula for calculating the amount of distortion of the entire image reflecting the characteristics of human vision by assigning a weight to each of the amount of distortion in the center region and the amount of distortion in the peripheral region.

In Equation 5, Dc represents a distortion amount of the center region and Dp represents a distortion amount of the peripheral region, respectively. Dc is multiplied by 1, and Dp is multiplied by w. In an embodiment of the present invention, 0.8 is substituted for the weight w. In this case, even if the same amount of distortion, the amount of distortion in the peripheral region is 0.8 times less than the amount of distortion in the central region, so that the amount of distortion in the central region will be relatively higher.

After the distortion of the entire image data is obtained, the PSNR may be obtained as shown in [Equation 6].

In Equation 6, M and N are the number of horizontal and vertical pixels on the screen.

The method of the present invention as described above can be written in a computer program. And the code and code segments constituting the program can be easily inferred by a computer programmer in the art. In addition, the created program is stored in a computer-readable recording medium (information storage medium), and read and executed by a computer to implement the method of the present invention. And the recording medium includes all types of recording media (intangible medium such as a carrier wave as well as tangible media such as CD and DVD) readable by a computer.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited by.

1 is a communication network configuration for explaining the operation of the image quality measuring apparatus according to the present invention.

2 is a configuration diagram showing a configuration of an image quality measuring apparatus according to the present invention.

3 is a flow chart showing the flow of the image quality measurement method according to an embodiment of the present invention.

4 is a histogram of each frequency component generated before dividing the prediction mode image into 4x4 blocks and performing discrete cosine transform and quantizing the DCT coefficients accordingly.

FIG. 5 is a histogram for each frequency component produced by dividing an intra mode image into 4 × 4 blocks and then performing discrete cosine transform and quantizing the DCT coefficients accordingly. FIG.

6 is a result of modeling a histogram for each frequency component of DCT coefficients using a Gaussian function.

7 shows a comparison of a histogram before quantization and a histogram after quantization of DCT coefficients.

8 is an enlarged view of a portion of FIG. 7 to explain a method of calculating quantization distortion. FIG.

9 is a diagram for describing a method of obtaining a Gaussian function value by dividing a quantization step interval into detailed intervals.

FIG. 10 is a diagram for describing a method of calculating a distortion amount of a quantization step section by calculating a distortion amount of each subdivision section. FIG.

FIG. 11 illustrates distortions of various images in order to experiment with how the human eye perceives differently according to the position of the screen even when the same amount of distortion occurs.

12 is a screen divided into a central region and a peripheral region in order to give a weight reflecting human visual characteristics.

13 is a flowchart illustrating a flow of a method for measuring image quality according to an embodiment of the present invention.

Claims (10)

Receiving image data; Obtaining discrete cosine transform coefficients and quantization parameters included in the image data; Generating a histogram for each frequency component of the image data using the discrete cosine transform coefficients; Generating a histogram modeling the histogram as a Gaussian function; And Calculating a quantization distortion amount for each frequency component of the image data by using the modeled histogram and the quantization parameter, The Gaussian function is defined as shown in the following [Equation 1].

The method of claim 1, K in the above [Formula 1] is defined as shown in [Formula 2].

The method of claim 1, Computing the quantization distortion amount for each frequency component Dividing the quantization step interval set in the modeled histogram into n subdivisions (where n is a positive integer); Calculating a function value for each subdivision of the quantization step interval using Equation 1; Calculating a distortion amount of the quantization step section by using the function value for each subdivision section and the following [Equation 3]; And Computing the quantization distortion amount for each frequency component by adding all the distortion amount of the quantization step interval Image quality measurement method comprising.

An input unit for receiving image data; A data manager obtaining a discrete cosine transform coefficient and a quantization parameter included in the image data; Generate a histogram for each frequency component of the image data by using the discrete cosine transform coefficients, generate a histogram of which the histogram is modeled by a Gaussian function, and use the modeled histogram and the quantization parameter to generate the histogram. An operation unit for calculating the quantization distortion amount for each frequency component of And the Gaussian function is defined as in Equation 1 below. The method of claim 4, wherein Wherein k in Equation 1 is defined as in Equation 2. The calculation unit The quantization step interval set in the modeled histogram is divided into n subdivisions (where n is a positive integer), and a function value for each subdivision period of the quantization step interval is calculated using Equation 1 above. And calculating a distortion amount of the quantization step section by using a function value for each subdivision section and [Equation 3], and calculating the quantization distortion amount for each frequency component by adding all the distortion amounts of the quantization step section. Receiving image data divided into a central area and a peripheral area; Calculating a quantization distortion amount for the center region and the peripheral region; Weighting the quantization distortion amount for the center region and the quantization distortion amount for the peripheral region, respectively; And Summing quantization distortion amount for each frequency component of the weighted center region and quantization distortion amount for each frequency component of the peripheral region. Image quality measurement method comprising. The method of claim 7, wherein Computing the quantization distortion amount for the center region and the peripheral region Obtaining discrete cosine transform coefficients and quantization parameters included in the image data; Generating a histogram for each frequency component of the image data using the discrete cosine transform coefficients; Generating a histogram modeling the histogram as a Gaussian function; And And calculating the amount of quantization distortion for each frequency component of the image data by using the modeled histogram and the quantization parameter, wherein the Gaussian function is defined as in Equation 1. The method of claim 8, In [Equation 1] K is an image quality measurement method defined as in Equation 2. The method of claim 8, Computing the quantization distortion amount for each frequency component Dividing the quantization step interval set in the modeled histogram into n subdivisions (where n is a positive integer); Calculating a function value for each subdivision of the quantization step interval using Equation 1; Calculating a distortion amount of the quantization step section by using the function value for each subdivision section and [Equation 3]; And Computing the quantization distortion amount for each frequency component by adding all the distortion amount of the quantization step interval Image quality measurement method comprising.

Images