KR100207392B1 - System for testing picture quality of reconstructed image and texture domain determinator therein - Google Patents

Abstract

According to the present invention, a reconstructed image capable of accurately evaluating the image quality of a reproduced image reconstructed from an encoded original image by using a signal-to-noise ratio level at a strong edge between the original image and the reconstructed reproduced image in consideration of human visual characteristics Relates to a picture quality evaluation system. The present invention first detects a texture area that is difficult to characterize because the content information is complicated in the original image and is very insensitive to human vision, and then, based on the detection result, the texture area in the original image and the reconstructed playback image using a mask. Remove it. Next, the contrast dependence between the original image and the reconstructed reproduced image from which the texture region is removed can be removed, and the image quality state of the reconstructed reproduced image can be accurately evaluated using the peak signal-to-noise ratio at the strong edge component between the two images. .

Therefore, since the present invention evaluates the image quality of the reconstructed reproduced image having only the strong edge component of the original image and the reconstructed reproduced image, which is very advantageous for the image quality evaluation, it is possible to accurately evaluate the image quality state of the reconstructed reproduced image by decoding the encoded image. Can be.

Description

Texture Area Judgment Device and Image Quality Evaluation System

1 is a block diagram of a system for evaluating image quality of a reconstructed image according to an exemplary embodiment of the present invention.

FIG. 2 is a block diagram of a texture area determination apparatus according to a preferred embodiment of the present invention employed in the image quality evaluation system of FIG.

3 shows N for determining the texture area according to the invention. Diagram showing an example of a mask used to remove coefficient values of the low frequency portion in a subblock of N. FIG.

4 illustrates N for determining a texture region according to the present invention. A diagram showing an example of performing projection in four directions on a N subblock.

5 is an exemplary view of a frame shown as an example for explaining the image quality evaluation process according to the present invention.

FIG. 6 is a diagram illustrating the intensity of each pixel value removed from the pixel average value in order to remove the dependence of brightness between the original image excluding the texture region and the reconstructed reproduced image according to the present invention. Fig. 6 (b) is a graph showing the contrast levels after removing the contrast dependencies.

* Explanation of symbols for main parts of the drawings

101: texture area determination block 102: low pass filter

103, 104: masking block 105, 106: average value calculation block

107,108 subtractor 110 PSNR calculation block

120, 1015, 1018: comparator 1011: split block

1012: DCT block 1013: low frequency masking block

1014: adder 1016: directional projection block

1017: average value calculation block 1019: AND gate

The present invention relates to a system for evaluating the quality of a reconstructed image decoded from a compressed coded image. More particularly, the present invention relates to a peak signal-to-noise ratio (PSNR) in a strong edge component, which is one of the main parts of an image. The present invention relates to a system for evaluating image quality between an original image and a reconstructed image before encoding.

As is well known in the art, the transmission of discrete video signals can maintain better image quality than analog signals. When a video signal composed of a series of image frames is represented in a digital form, a considerable amount of data must be transmitted, especially in the case of high-definition television (HDTV). However, since the usable frequency range of the conventional transmission channel is limited, in order to transmit a large amount of digital data, it is necessary to compress the data to be transmitted and reduce the amount of transmission. In addition, the compressed video signal and the audio signal are encoded through different coding techniques due to the characteristics of those signals. In this encoding, a video signal compression technique that generates a greater amount of digital data than the audio signal is particularly important. It can be said to take part.

On the other hand, as the various compression techniques mainly used for encoding the image signal, a hybrid encoding technique combining a stochastic encoding technique and a temporal and spatial compression technique is known to be the most efficient.

Most hybrid coding schemes, which are one of the efficient coding schemes described above, use motion compensated DPCM (differential pulse code modulation), two-dimensional discrete cosine transform (DCT), quantization of DCT coefficients, VLC (variable length coding), and the like. Here, the motion compensation DPCM determines a motion of the object between the current frame and the previous frame, and predicts the current frame according to the motion of the object to generate a differential signal representing the difference between the current frame and the prediction value. Such methods are described, for example, in Fixed and Adaptive Predictors for Hybrid Predictive / Transform Coding by Staffan Ericsson, IEEE Transactions on Communication, COM-33, NO.12 (1985, December), or Amotion Compensated Interframe Coding by Ninomiy and Ohtsuka. Scheme for Television Pictures, IEEE Transactions on Communication, COM-30, No. 1 (January, 1982).

More specifically, the motion compensation DPCM predicts the current frame from the previous frame according to the motion of the object estimated between the current frame and the previous frame. Here, the estimated motion may be represented by a two-dimensional motion vector representing the displacement between the previous frame and the current frame.

In general, there are various approaches to estimating the pixel displacement of an object, and these are generally classified into two types, one of which is a block-based motion estimation method and the other is a pixel-based motion estimation method. In the estimation, the block of the current frame is compared with the blocks of the previous frame to determine an optimal matching block, and then, from this, the interframe displacement vector (how much the block moves between frames) for the entire current frame is estimated. do.

Therefore, when transmitting a video signal, the transmitting side compresses and encodes the image signal in the buffer of the output side in order by taking into account the spatial and temporal correlation of the image signal in block units or pixel units through the above-described encoding technique. The image data will be transmitted to the decoding system at the receiving side via the transmission channel at a desired bit rate in response to the channel request.

More specifically, the encoding system on the transmitting side removes spatial redundancy of the video signal by using transform coding such as discrete cosine transform (DCT), and further uses temporal encoding of the video signal by using differential coding through motion estimation and prediction. By eliminating redundant redundancy, the video signal can be efficiently compressed.

Therefore, in the decoding system on the receiving side, the video signal encoded through the reverse process (variable decoding, inverse quantization, IDCT, etc.) in the transmitting encoding system as described above for display on the monitor is converted into the previous original image signal. Will be restored.

The DPCM / DCT hybrid coding scheme described above has a target bit rate of Mbps, and its application fields are CD-ROM, computers, home appliances (digital VCRs, etc.), broadcasting (HDTV), and the like. It is mainly concerned with MPEG-1,2 and H.261 encoding algorithms for high bit rate encoding, which mainly consider only the statistical characteristics of block-by-block motion in the image, which has already been completed by the standard.

On the other hand, in recent years, due to the improvement and spread of PC performance, the development of digital transmission technology, the realization of high-definition display device, the development of memory device, various devices such as home appliances can process and provide image information with huge data. In order to meet this demand, the transmission of audio-video data through the existing low-speed transmission path (PSTN, LAN, mobile network, etc.) and the limited capacity storage device to meet these demands. New storage techniques with high compression rates are needed for storage.

However, the existing video encoding techniques as described above are based on local block motion in the entire image regardless of the shape of the moving object and the global motion. Therefore, the existing video coding techniques (hybrid coding systems using MC-DCT) are applied to the playback image where the deterioration of image quality such as blocking, corner blur, and spot phenomenon is finally restored when applying block-by-block motion compensation coding at a low data rate. Will appear. In addition, in order to maintain the resolution for low data transmission, high compression ratio of image data is required, which cannot be implemented by an existing DCT transform based coding scheme having a compression ratio of about 40: 1.

Therefore, there is a need for a coding method standard for realizing additional compression of about 4: 1, compared to the existing coding method based on the DCT transform, and subjective picture quality is based on recent human visual characteristics according to the needs of the times. The research on low bit rate video encoding technique for the development of MPEG4 standard, which focuses on the high quality, has been actively conducted everywhere.

In order to meet these needs, potential viable low bit rate video coding techniques are currently being studied, such as Wave-Based Coding and Model-Based Coding to improve existing coding techniques. Object-Based Coding, a kind of coding, Segmentation-Based Coding that divides and encodes an image into a plurality of subblocks, and Fractal Coding that uses self-similarity of an image. There is this.

As is well known in the art, object-by-object coding divides an image into regions with similar pixel values (for example, objects and backgrounds), and then separates the contours representing the boundaries and the content information of the region (object). It is a method of encoding. Here, contour detection of an object is achieved by detecting edges using the brightness distribution of the original image, as is well known in the art. However, since the contour and the content information are completely different information, their encoding methods are independent of each other. Therefore, at the time of encoding, the contour and its brightness information are separated and encoded, and then multiplexed and transmitted through a multiplexer.

On the other hand, the coding system and decoding system employing the hybrid coding technique using the MC-DCT as described above, or the coding system and decoding system employing the MPEG-4 related low rate coding technique that emphasizes subjective picture quality based on human visual characteristics. In this case, the researchers of each affiliated company (developer's research institute, etc.) conduct various tests such as the desired coding rate and coding degree of the image data through numerous simulations at the initial stage of system development and improvement after completion of basic development. In addition, various tests (e.g., image quality evaluation of the reconstructed image) are executed for the process of reconstructing the compressed-coded image to the original signal. Herein, the present invention relates to the improvement of a system for evaluating the image quality of a reproduced image, which is obtained by reconstructing a compression-coded image into an original signal, in particular during such simulation.

In the conventional method of evaluating the image quality of a reproduced image, the image quality evaluation of the reconstructed reproduced image is performed by comparing the overall peak signal-to-noise ratio (PSNR) between the original image and the reconstructed reproduced image before encoding. In other words, the compression-encoded video is reconstructed into the original signal before being coded by the decoding system, and then the PSNR between the reconstructed reproduced video and the original video is obtained. The researcher will be able to evaluate the image quality of the reconstructed playback image based on the evaluation of the image quality of the reconstructed playback image, that is, whether the PSNR result exceeds the preset value. In this case, the predetermined value may be variably set by the researcher according to a field to which a coding technique is applied, for example, a field to which a hybrid coding technique using MC-DCT is applied or a field to which a low bit rate coding technique is applied. . As an example, the inventor of the present invention sets a reference value (preset value) of the PSNR between the reproduced video and the original video at about 25 dB during the evaluation of the image quality of the reconstructed reproduced video for the compression-coded video by using a low bit rate encoding technique. The simulation was performed.

Therefore, the image quality evaluation system as described above may be effectively used for improving the coding efficiency or decoding efficiency of a desired video.

On the other hand, as described above, a typical conventional method of evaluating the image quality of a reproduced image using the overall PSNR between the original image and the reconstructed reproduced image is based on the PSNR of the entire image without considering human visual characteristics. In addition to evaluating image quality, i.e., it is very insensitive to human visual characteristics and its contents are so complex that it is difficult to analyze statistical characteristics, for example, it is filled with diagonal lines in the image as shown in FIG. By including all of the parts indicated by b), there is a problem in that the reliability and precision of image quality evaluation of the reproduced video are inferior. In other words, most of the texture information is a very complex part of the content information (texture information) included in edges or contours, that is, the texture of the tree forest, and such complex texture information can be described by a statistical model. The disadvantage is that the characteristics are very difficult.

Accordingly, the present invention solves the problems of the prior art as described above, and is used to evaluate the image quality state of the reconstructed reproduced image using the peak signal-to-noise ratio between the reproduced image between the original image and the reconstructed reproduced image. It is an object of the present invention to provide a device capable of determining possible texture areas.

Another object of the present invention is to use the texture region determination device, and to consider the visual characteristics of the human being, using the signal-to-noise ratio level at the strong edge between the original image and the reconstructed reproduced image, The present invention provides a system for evaluating image quality of reconstructed images that can be accurately evaluated.

In accordance with an aspect of the present invention, there is provided an apparatus for determining a texture area in a system for evaluating a picture quality state of a reconstructed reproduced image by using a peak signal-to-noise ratio between an uncoded original image and a reconstructed reproduced image. N, wherein the original video or the restored playback video An image segmentation block divided into N subblocks; The divided N N-time image signal of N subblocks using cosine function A DCT block for converting to a DCT transform coefficient value in a frequency domain of N; N above Low-pass coefficient removing means for removing N transform coefficients distributed in the lowest frequency portion using a low-frequency mask for each DCT transform coefficient block of N; N above A comparator for comparing an absolute total value of the remaining transform coefficients except the N transform coefficients removed from the DCT transform coefficient block of N with a preset threshold Th1 to generate an output signal corresponding to the comparison result; N above For each DCT transform coefficient block of N, N one-dimensional data is generated through projection in multiple directions, and average values for each N directional coefficient values belonging to each generated one-dimensional data are respectively calculated. Means for calculating difference values between respective average values; By comparing the calculated difference values with a predetermined threshold value Th2 and checking whether there is a difference value larger than the threshold value Th2 among the difference values, the N Directional edge detection means for detecting a directional edge component for each DCT transform coefficient block of N; And N in the image quality evaluation system based on the output control signal from the comparator and the directional edge detection signal from the directional edge detection means. A texture area determining apparatus in a picture quality evaluation system comprising decision information generating means for providing texture area decision information for each DCT transform coefficient block of N is provided.

According to another aspect of the present invention, there is provided a system for evaluating an image quality state of a reconstructed reproduced image by using a peak signal-to-noise ratio between an unencoded original image and a reconstructed reproduced image. N A texture region determination block which divides into subblocks of N and then detects whether or not the texture region is divided in units of the divided subblocks and generates determination information corresponding to the detection result; The divided N according to the texture area determination information N corresponding to N subblock A first masking block to remove the texture area from the original image by masking using an N mask; Compute an average value of pixel values of the original image from which the texture area is removed, and subtract each pixel value of the original image from which the texture area is removed by the calculated average pixel value, respectively, to remove the texture area. First means for generating a; The N according to the texture area determination information Another N corresponding to N mask A second masking block for removing the texture area from the reconstructed reproduced image by masking using an N mask; Calculating an average value of pixel values of the reconstructed reproduced image from which the texture region is removed, and subtracting each pixel value of the reconstructed reproduced image from which the texture region is removed by the calculated average pixel value, respectively, to remove the texture region; Second means for generating a new restored restored image; A peak signal-to-noise ratio of the new original image from which the texture region has been removed and a peak signal-to-noise ratio of the new reconstructed reproduced image from which the texture region is removed, respectively, and a peak that calculates a difference between the calculated peak signal-to-consonant ratios; A signal-to-noise ratio calculation block; And a comparator configured to compare the difference between the calculated peak signal-to-noise ratios and a preset threshold Th3, and generate a result signal for evaluating an image quality state of the reconstructed reproduced image corresponding to the comparison result. Provide a picture quality evaluation system.

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

First of all, the present invention implements the image quality between the original image and the reconstructed reproduced image by excluding texture information relatively insensitive to human visual characteristics and having only PSNR information in the strong edge portion, which is one of the main display information of the image. Evaluation, that is, the distortion degree of the reproduced image obtained by reconstructing the encoded image is evaluated.

Accordingly, the present invention employs a technical means for evaluating the distortion degree of the reconstructed reproduced image by referring to only PSNR information in the strong edge portion, so that most of the above-described conventional methods are very complicated at the time of image quality evaluation. It is possible to remarkably improve the accurate image quality evaluation and the degradation of reliability caused by the texture information. In particular, the present invention using the edge component strong in image quality evaluation of the reconstructed image is superior to the image coded by the low-rate image coding technique such as object-based encoding technique rather than the image coded by the hybrid coding technique employing the MC-DCT. It will provide picture quality evaluation characteristics.

1 is a block diagram of a system for evaluating image quality of a reconstructed image according to an exemplary embodiment of the present invention. As shown in the drawing, the image quality evaluation system of the present invention includes a texture region determination block 101, a first masking block 103, and a first average value calculating block constituting a path for signal processing an unencoded original image. 105, a low pass filter (LPF) 102, a second masking block 104, and a second average value calculating block forming a path for signal-processing the reproduced video obtained by reconstructing the compressed coded video with the first subtractor 107. 106, a second subtractor 108, a PSNR calculation block 110 and a comparator 120 connected to the output side of the two subtractors 107, 108.

In FIG. 1, the texture region determination block 101 detects an area which is very insensitive to human visual characteristics, that is, a statistical characteristic that is very complex and difficult to analyze in the original image provided from the input side. The detailed operation process of determining the texture region in the original image will be described in detail below with reference to FIG. 2 showing the detailed internal structure thereof.

That is, FIG. 2 is a detailed block diagram of the texture region determination block 101 described above. As shown in the drawing, the texture region determination block 101 includes a partition block 1011, a DCT block 1012, A low frequency masking block 1013, an adder 1014, a first comparator 1015, a directional projection block 1016, an average value calculation block 1017, a second comparator 1018 and an end gate 1019.

Referring to FIG. 2, the splitting block 1011 receives N uncoded original video signals inputted through line L11. A subblock of N (eg 4 N subdivided into four subblocks) The N subblocks are provided in the next DCT block 1012. In this case, subblock segmentation of the original image is performed according to the size of the mask used to remove the texture area. 8 or 16 It can be divided into blocks such as 16, but considering the search efficiency of the texture area in the whole image 4 It is most preferable to divide into four subblocks.

On the other hand, in the DCT block 1012, N provided from the above-described partition block 1011. A video signal (pixel data) in a time domain in units of N subblocks is obtained by using a cosine function. Convert to DCT conversion coefficients in the frequency domain in units of N subblocks. Then, in this way N The DCT transform coefficients converted in units of N subblocks are provided to the low frequency masking block 1013 and the directional projection block 1016 through the line L22, respectively.

At this time, DCT block 1012 N The corresponding N output from the N subblock If the N subblock is a texture region, the spatial frequency is high, so the coefficients will be distributed in the high frequency part. Accordingly, in the low frequency masking block 1013, to examine how the coefficients are distributed in the high frequency portion except for the low frequency portion, a corresponding N is used by using the mask shown in FIG. Masking is performed on the N subblocks, and as a result, coefficients (three transform coefficients) distributed in the low frequency portion as indicated by hatching in FIG. 3 (b) will be removed.

Next, the adder 1014 adds all the coefficient values except for the three coefficients of the low frequency portion to calculate an absolute total value, and the absolute total value thus calculated is the first comparator 1015 of the next stage. Is provided. Accordingly, the first comparator 1015 compares the absolute total value provided from the adder 1014 with a predetermined threshold Th1, and generates an output signal corresponding to the comparison result to generate one side of the AND gate 1019. To the terminal. That is, the first comparator 1015 generates an output signal of logic 1 when the absolute total value is greater than the preset threshold Th1 and, on the contrary, generates an output signal of logic 0 when it is smaller than the preset threshold Th1. do.

On the other hand, in the directional projection block 1016, N input from the above-described DCT block via line L22. For each sub-block of N, for example, as shown in FIG. 4 (b), respectively, four projections are generated in four directions to provide the average value calculation block 1017. do. That is, the directional projection block 1016 has the first one-dimensional data (coefficients a1, a2, 4, 4) through the projection in the horizontal direction as shown in (a) of FIG. 4 (b). a3, a4), and the second one-dimensional data (coefficients a1, a5, a9, a9, through the projection in the vertical direction as shown in (b) of FIG. 4 (b)). a13), and the third one-dimensional data (coefficients a1, a6, a11, a16 of FIG. 4 (a)) through diagonal projection as shown in (c) of FIG. 4 (b). Then, the fourth one-dimensional data (coefficients a4, a7, a10, a13 of FIG. 4 (a)) are generated by projecting in a different diagonal direction as shown in (d) of FIG. 4 (b). Create

Next, the average value calculation block 1017 calculates the degree of change for the four one-dimensional data. That is, the average values for each of the four coefficient values belonging to each one-dimensional data are respectively calculated and provided to the second comparator 1018 of the next stage.

On the other hand, the second comparator 1018 checks whether there is a difference value larger than the predetermined threshold value Th2 among the difference values (degrees of change) between the average values provided from the above average value calculation block 1017. In other words, the second comparator 1018 compares the difference values between the average values and the preset threshold value Th2, respectively, and generates an output signal corresponding to the comparison result, and provides it to the other terminal of the AND gate 1019. do. That is, the second comparator 1018 generates an output signal of logic 0 when any one of the difference values between the average values is larger than the preset threshold Th2, and conversely, the second comparator 1018 is larger than the preset threshold Th2. If not present, it generates an output signal of logic 1.

As mentioned above, N The reason for performing the difference value comparison between the average values of the four one-dimensional data obtained by projecting in four directions with respect to each subblock of N is that the high frequency component in the adder 1014 described above is used when each subblock is not a texture area. This is to check that there is a strong directional edge for the oversubblocks because it cannot exceed the absolute sum of. Thus, such subblocks (subblocks having strong directional edges) are not determined as texture regions.

Thus, the AND gate 1019 determines whether or not each subblock is a texture area based on logic signals provided from the two comparators 1015 and 1018 as described above, for example, the corresponding subblock. A logic signal of 1 is provided to the masking blocks 103 and 104 of FIG. 1 through a line L13 when a logic signal of 1 is used in this texture area and a logic signal of 0 when the corresponding subblock is not a texture area.

Through the process as described above, in the texture area determination block 101 of FIG. A determination signal as to whether each subblock of N is a texture area is provided to the first and second masking blocks 103 and 104 via the line L13, respectively.

Referring back to FIG. 1, the first masking block 103 is divided by N in the division block 1011 of FIG. 2 in accordance with the determination signal from the texture area determination block 101 as described above. N corresponding to N subblock The image of the portion determined as the texture area is removed from the original image signal on the line L11 using the N mask. As an example, as shown in FIG. 5, if the texture area in the input original image is hatched and indicated by reference numeral b, this texture area will be removed through masking. In FIG. 5, reference numeral a denotes an edge portion, and (c) denotes a background portion. That is, the masking according to the present invention removes the test area (a forest of a tree or a spectator moving on a playground), which is a very insensitive part of the human vision, and is very effective in realizing the background and high-quality image evaluation. Only advantageous strong edge components will remain.

On the other hand, the reconstructed reproduced video signal on the line L12 is filtered through the low pass filter 102 and then provided to the second masking block 104. Here, the reason for low-pass filtering the reconstructed reconstructed video is that the reconstructed video typically removes a lot of block noise that appears to be very annoying to human vision due to quantization during encoding, in particular, block noise that appears in the flat portion of the video. to be.

Accordingly, in the second masking block 104, N is based on the determination signal provided from the texture area determination block 101 described above via the line L13. N corresponding to N subblock An image of a portion determined as a texture area is removed from the reconstructed reproduced video signal on the line L12 using the N mask. That is, as in the above-described first masking block 103, as shown in FIG. 5, if the texture area in the inputted reconstructed reproduced image is filled with hatched and indicated by reference numeral b, The texture area will be removed by masking.

Here, the second masking block applies the texture region determination information from the texture region determination block 101 as described above to the masking, so that the original image and the reconstructed reproduced image for the substantially same image are the edge portion, the texture. This is because the regions and the like can be regarded as being substantially the same with almost no difference. That is, it can be considered to prevent the addition of unnecessary hardware (hardware for determining the texture area of the restored reproduced video). In addition, as in the first masking block 103 described above, a texture region (a forest of a tree or a spectator moving on a playground), which is a very insensitive part of human vision, in the reproduced image restored through masking according to the present invention, Only a strong edge component will be left out, which is eliminated and very advantageous for its background and for the realization of high quality image evaluation.

On the other hand, the original image from which the texture region (Fig. 5 (b)) is removed and the reconstructed reproduced image through the process as described above, as shown in Fig. 6 (a) as an example, the difference in brightness (brightness) In the present invention, an average pixel value of the remaining portions of each image (original image and reproduced image) from which the texture area is removed is removed in order to remove the contrast dependency, which may be a deterrent to high-quality image evaluation. Is calculated and then converted into an image (original and playback video) signal obtained by subtracting the average pixel value from each pixel value. In FIG. 6A, the portion indicated by the solid line represents the contrast graph of the original image, and the portion indicated by the dotted line represents the contrast graph of the restored playback image.

To this end, the original image (part (c) of FIG. 5) from which the texture area is removed through the first masking block 103 is provided to the first average value calculating block 105, and the first average value calculating block 105 An average value of pixels of the original image from which the texture area is removed is calculated and provided to the first subtractor 107.

Therefore, in the first subtractor 107, the average pixel value from the first average value calculation block 105 for each pixel value of the original image from which the texture area provided from the first masking block 103 is removed. Are respectively subtracted and then provided to the next stage PSNR calculation block 110 through the line L15.

On the other hand, as described above, the reconstructed reproduced image (part (c) of FIG. 5) in which the texture region is removed through the second masking block 104 is provided to the second average value calculating block 106, and the second average value calculating block In operation 106, an average value of pixels of the reproduced image from which the texture area is removed is calculated and provided to the second subtractor 108.

Accordingly, in the second subtractor 108, the average from the second average value calculation block 106 is applied to the respective pixel values of the reconstructed reproduced image from which the texture area provided from the second masking block 104 described above is removed. Each pixel value is subtracted and then provided to the next stage PSNR calculation block 110 through the line L14.

For example, as shown in FIG. 6 (a), even if there is a difference in contrast that may occur between the original image and the reconstructed reproduced image, which may be a detrimental factor in the accurate image quality evaluation, through the above-described process By subtracting the pixel value of the portion where the texture area has been removed by the average pixel value, as shown in FIG. 6B, the difference in contrast between the original image and the reconstructed reproduced image can be significantly reduced. Therefore, the present invention enables the implementation of high-definition picture quality evaluation on the reconstructed reproduced video. In FIG. 6B, the portion indicated by the solid line represents the intensity graph of the original image, and the portion indicated by the dotted line represents the intensity graph of the reconstructed reproduced image.

On the other hand, the PSNR calculation block 110 obtains the peak signal-to-noise ratio (PSNR) of each of the original image and the reconstructed reproduced image, each of which has been removed from the contrast dependency, provided through the respective lines L15 and L14. The difference between each obtained PSNR (PSNR of the original image and PSNR of the reproduced image) is calculated, and the difference value between each PSNR calculated in this way is provided to the comparator 120 of the next stage. Here, the unit of PSNR is dB (decibel) commonly used in the art.

Accordingly, the comparator 120 compares the difference between the two PSNRs provided from the PSNR calculation block 110 and the preset threshold Th3, and compares the output signal corresponding to the comparison result, that is, the reconstructed playback image. A control signal (logic 1 or logic 0) for evaluating image quality is provided to a controller (e.g., a microprocessor) of the entire evaluation system, not shown.

In this case, the predetermined value is variable by the researcher (or worker) depending on the field to which the coding technique is applied, for example, the field to which the hybrid coding technique using MC-DCT is applied or the field to which the low bit rate coding technique is applied. Will be set. As an example, the inventor of the present invention sets a reference value (preset value) of the PSNR between the reproduced image and the original image at about 25 dB when the image quality of the reconstructed reproduced image of the compressed coded image is low through the low power rate encoding technique. The simulation was performed.

Therefore, the determination signal for the evaluation of the image quality of the reproduced image visually or acoustically through an unillustrated display means driven according to the output control signal from the microprocessor based on the control signal provided from the comparator 120. Is displayed, the operator (or researcher) will be able to grasp the high-definition image quality of the image currently restored and reproduced.

As described above, according to the texture region determination apparatus of the present invention and the image quality evaluation system for reconstructed images using the same, strong edges of the original image and the reconstructed reproduced image are very advantageous for image quality evaluation except for the texture region whose statistical characteristics are complicated. By evaluating the image quality of the reconstructed reproduced video only with the components, it is possible to accurately evaluate the image quality state of the reproduced reproduced video by decoding the encoded image.

Claims (8)

A texture region determination apparatus in a system for evaluating the image quality state of the reconstructed reproduced image by using a peak signal-to-noise ratio between the original image and the reproduced reproduced image from the encoded original image, wherein N An image partition block divided into N subblocks; The divided N N-time image signal of N subblocks using cosine function A DCT block for converting to a DCT transform coefficient value in a frequency domain of N; N above Low-pass coefficient removing means for removing N transform coefficients distributed in the lowest frequency portion using a low-frequency mask for each DCT transform coefficient block of N; N above A comparator for comparing an absolute total value of the remaining transform coefficients except for the N transform coefficients removed from the DCT transform coefficient block of N with a preset first threshold value Th1 to generate an output signal corresponding to the comparison result; N above For each DCT transform coefficient block of N, N one-dimensional data is generated through projection in multiple directions, and average values for each N directional coefficient values belonging to each generated one-dimensional data are respectively calculated. Average value calculating means for calculating difference values between respective average values; By comparing each difference value calculated by the average value calculating means with a preset second threshold value Th2 and checking whether there is a difference value larger than the threshold value Th2 among the difference values, the N Directional edge detection means for detecting a directional edge component for each DCT transform coefficient block of N; The N to the image quality evaluation system based on the output control signal from the comparator and the directional edge detection signal from the directional edge detection means. And determination information generating means for providing texture region determination information for each DCT transform coefficient block of N. The apparatus of claim 1, wherein the directional coefficient values belonging to the generated one-dimensional data are N directional coefficient values. The apparatus of claim 1, wherein the low-pass coefficient removing unit removes three transform coefficients of the lowest frequency part adjacent to each other. The said average value calculation means is a said N, And the four-dimensional data is generated by projecting each DCT transform coefficient block of N in four directions. 5. The apparatus of claim 4, wherein the four projection directions are horizontal, vertical, and two diagonal directions facing each other. A system for evaluating an image quality state of a reconstructed reproduced image by using a peak signal to noise ratio between an original image and a reproduced image reconstructed from an encoded original image, wherein the original image is N. A texture region determination block which divides into subblocks of N and then detects whether or not the texture region is divided in units of the divided subblocks and generates determination information corresponding to the detection result; The divided N according to the texture area determination information N corresponding to N subblock A first masking block to remove the texture area from the original image by masking using an N mask; Compute an average value of pixel values of the original image from which the texture area is removed, and subtract each pixel value of the original image from which the texture area is removed by the calculated average pixel value, respectively, to remove the texture area. First means for generating a; The N according to the texture area determination information Another N corresponding to N mask A second masking block for removing the texture area from the reconstructed reproduced image by masking using an N mask; Calculating an average value of pixel values of the reconstructed reproduced image from which the texture region is removed, and subtracting each pixel value of the reconstructed reproduced image from which the texture region is removed by the calculated average pixel value, respectively, to remove the texture region; Second means for generating a new restored restored image; The peak signal to noise ratio of the new original image from which the texture region has been removed and the peak signal to noise ratio of the new reconstructed reproduced image from which the texture region has been removed are calculated, respectively, and the peak signal calculating the difference between the calculated peak signal to noise ratio Noise-to-noise ratio calculation block; And comparing a difference between the calculated peak signal-to-noise ratios and a preset threshold value Th3, and generating a result signal for evaluating an image quality state of the reconstructed reproduced image corresponding to the comparison result. Image quality evaluation system of restored image. 7. The image quality of the reconstructed image as claimed in claim 6, wherein the first and second means remove the contrast dependency between the new original image from which the texture region is removed and the new reconstructed reproduction image from which the texture region is removed. Evaluation system. The system of claim 6 or 7, wherein the predetermined threshold (Th3) is variably set according to an application of encoding and decoding of the original image.