KR100212823B1 - Blocking phenomenon removal device and video decoder using the same - Google Patents

Abstract

The present invention relates to a block removing device and a video decoder using the block removing device, wherein the block removing device includes a plurality of low pass filters, a threshold setting unit and a comparison selection unit, and the low pass filtered image signal and the pixel value of the original image. The low pass filter results only when the difference is less than the threshold value as the output of the blocking phenomenon elimination device, thereby eliminating the blocking phenomenon without blurring the minute portion or the outline of the original image, and the threshold value used here is that the pixel to be processed is flat. The value changes depending on whether you are located in an area or in a complex area, which has the advantage of better blocking in flat areas and better conservation of contours in complex areas (represented in Figure 3). ), And also to remove the blocking of the picture at the output of the existing video decoder By providing an additional device for removing blocking artifacts group, to remove the blocking artifacts caused by the quantization noise of the DC coefficient has the advantage that can obtain images of further improved image quality.

Description

Blocking phenomenon removal device and video decoder using the same

1 is a block diagram showing the configuration of a general video encoder.

2 is a block diagram showing the configuration of a general video decoder.

Figure 3 is a detailed block diagram showing the configuration of the blocking phenomenon removal apparatus according to the present invention.

4 is a block diagram showing the configuration of a video decoder using the block phenomenon removing apparatus according to the present invention.

* Explanation of symbols for main parts of the drawings

50: plurality of low pass filter 52: regional dispersion acquisition unit

54: first comparator 56: first switch

58: adder 60: second comparator

62: subtractor 64: third comparator

66: selection signal generator 68: second switch

80: buffer 82: variable length decoder

84: inverse quantizer 86: inverse discrete cosine transform unit

88: frame memory 90: motion compensation unit

92: Adder 94: Blocking phenomenon removal device

100: video encoder 200: threshold setting unit

300: comparison selection

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a blockage removing apparatus and a video decoder using the same. In particular, a blockage removing apparatus for removing a blocking phenomenon due to quantization noise of a transform-coded DC coefficient to obtain a further improved image quality, and the same The video decoder used.

In general, HDTV (high definition television) has more than twice the vertical and horizontal resolution of current television, improves the problems of current television such as cross color, and has an aspect ratio of 16: 9 that is wider than current television. Compact Disc A television with digital audio performance at the sound quality level.

The HDTV system developed so far is divided into two types based on the transmission method. The first is a sample value transmission method in which a digital signal is regarded as an analog signal and transmitted. MUSE and HD-MAC of Japan correspond to this method. The second is a fully digital method that digitalizes from signal processing to transmission, and is currently rapidly reaching the practical stage, starting with the United States. In particular, the full digital scheme is used by current broadcasters. The number of existing broadcast channels is increased by 4-10 times in frequency bandwidth. Compared to conventional color televisions, HDTV signals having twice the resolution of the vertical and horizontal axes and four times the resolution overall have too much bandwidth to transmit using the existing limited channels. Therefore, band compression is essential for the transmission of HDTV signals.

Among the HDTV systems classified according to the transmission method, the following will be described based on the fully digital method in the US.

Three methods are used in the fully digital HDTV system for band compression. First, the method of reducing the amount of transmission information by using the relationship between the image and the next image, that is, the motion vector, and secondly, compressing the information in one image. Discrete cosine transform and quantization technique, and thirdly, Variable Length Coding (VLC) technique converts the transformed information into bits of appropriate size by stochastic distribution of information.

6 In the current NTSC terrestrial broadcasting channel with a bandwidth of about 20Mbps, this channel capacity is about 20Mbps, and the video signal occupies about 17Mbps, and the audio signal occupies 0.5Mbps, and the remaining portion occupies other auxiliary data, text broadcasting, error correction code, etc. Done. Therefore, in simple PCM (Pulse Code Modulation), the data amount of video signal up to 1.2Gbps should be compressed and transmitted to 70: 1 or more.In order to minimize the loss of video information during compression, the redundancy inherent in video signal is effectively removed. Should be. Redundancy inherent in an image signal includes redundancy between color components, temporal redundancy in the time axis direction between the screen and the screen, and spatial redundancy between neighboring pixels in the screen space.

1 is a block diagram of a motion compensated hybrid encoder that is a general video signal data compression method widely used not only for full digital HDTV but also for various applications, and uses transform coding such as discrete cosine transform (DCT) to remove spatial redundancy in the image. Motion compensation differential image protection is used to remove temporal redundancy. In addition, variable length coding (VLC) is used to remove redundancy of the data bit stream itself.

Referring to FIG. 1, the subtractor 10 generates a difference image between frames by subtracting a motion compensation image of an original image of a current frame and a reconstructed image of a previous frame, which are input in units of frames. The discrete cosine transforming unit 12 outputs a discrete cosine change coefficient by performing discrete cosine transforming of the inter-frame difference image or the input image into blocks of 8 * 8 units to remove the correlation between pixels. The quantizer 14 quantizes the coefficients of the discrete cosine difference of the inter-frame difference image or the input image output from the discrete cosine varying unit 12 at a predetermined quantization interval and outputs the quantized interval. The variable length encoder 16 performs variable length protection on the inter-frame difference image or the input image quantized by the quantizer 14. Since the length of the data output from the variable length encoder 16 is not constant, the buffer 18 temporarily stores the data and outputs the data at a constant speed. In this case, the buffer 18 performs a quantization step according to the occupancy of the data. The size is determined to control the quantizer 14.

The inverse quantizer 20 is connected to the output terminal of the quantizer 14 and restores the quantized image to a state before being input to the quantizer 14. An inverse discrete cosine transformer 22 is connected to an output terminal of the inverse quantizer 20, and the inverse quantized inter-frame difference image or input image that is inversely quantized by the inverse quantizer 20 is transferred to the discrete cosine transformer 12. Restore to the state before input. The adder 24 adds the inter-frame difference image or the input image and the motion compensation image reconstructed by the inverse discrete cosine transform unit 22 to store the reconstructed image of the previous frame in the frame memory 26. The motion estimation unit 28 estimates a similar portion between the input image of the current frame and the reconstruction image of the previous frame stored in the frame memory 26 and outputs the result of the position movement as a motion vector. The motion compensator 30 outputs, to the subtractor 10 and the adder 24, the motion compensation image compensated for by the motion vector of the motion position of the reconstructed image of the previous frame stored in the frame memory 26. .

A process of encoding an image with reference to FIG. 1 will now be described in detail.

The R, G, and B signals output from the camera are highly correlated among the color components, and are converted to Y, U, and V color systems to reduce the correlation and match the human visual characteristics. The most sensitive Y component in human eyes is about 30 Occupies the band of U but V is about 15 Since it only takes a degree, there is almost no loss of information even if the number of samples is reduced by decimation. In general, Y: U: V is often 4: 1: 1, and especially in interlaced signals, all color signals of even (or odd) fields are lost by vertical decimation. The 4: 2: 0 format is also used. Also, because the bandwidth of the V signal is slightly wider than U, artificial artifacts can be prominent in either the red or blue line if the data allocation of V and U is not well balanced.

In order to eliminate temporal redundancy by using the correlation between neighboring screens, a motion compensated differential PCM (DPCM) scheme is used. The screen is divided into macroblocks of a constant size to obtain a motion vector indicating where each macroblock has moved from before, and motion compensation is performed using this vector.

The search range and precision of the motion vector change the magnitude of the error signal, which changes the amount of final coded data. In most images, since the horizontal motion is larger than the vertical direction, the search area is usually wider in the horizontal direction, and the search area is determined by the maximum displacement between neighboring screens.

The precision of motion vectors is often up to pixel units or even more precise half-pixel units. A commonly used method for estimating a motion pixel in half-pixel units is to first obtain a motion vector in pixel units by a block matching algorithm (BMA) or the like, and then obtain values of half-pixel positions around the vector by interpolation. A second full-search is performed based on nine half-pixel unit vector positions including the pixel-by-pixel motion vector, and the position having the least error is determined as the half-pixel unit motion vector.

On the other hand, in the interlaced scanning video signal, there may be a motion between two fields in one frame, so that motion compensation in units of fields may be more effective. In the frame-by-frame motion compensation, one macroblock needs one motion vector, but in the field-by-field motion compensation, one macroblock has two motion vectors. Combining the two methods and using an adaptive method that takes the more efficient motion compensation method among the two macroblocks, the complexity increases, but the data amount can be further reduced.

Since spatial redundancy has a high correlation between adjacent pixels in the screen, it is removed through a discrete cosine transform (DCT) process, which is a type of transform coding in block units (8 * 8). Fourier transforms are used to convert time-based signals to frequency-based signals. The discrete cosine transform (DCT) also converts pixel values into time-dependent signals based on matrix operations. The high autocorrelation of an image means that there are many parts that change slowly, and it can be considered that energy is concentrated in low frequency components in frequency. Discrete Cosine Transform (DCT) is based on the large correlation of the video signal in the spatial direction.The advantage of these methods is that the energy dispersed in all the pixels of the image is concentrated on several transform coefficients with low frequency including DC component. It is a method using appropriately. Discrete cosine transform (DCT) is performed in units of 8 * 8 blocks, and inter-image prediction error (inter-picture prediction error) when removing spatial redundancy in the intraframe coding method and interframe coding method. Is used to remove the correlation in the spatial direction. Since the energy is concentrated in a low frequency component, the quantization of the direct current component displayed in the upper left corner of one block 8 * 8 is performed in fine detail. As the high frequency component is used, sparse quantization can reduce an error caused by quantization. However, since the discrete cosine transform (DCT) is coded on a block basis, the block boundary is likely to be discontinuous, and the boundary is easily visible. If there is an edge in the block, a high frequency component appears, but if the quantization is sparse, the quantization error is ringing and widened into the block, which is perceived as image quality degradation such as mosquite noise or corona effect.

After a video signal is transformed into a frequency domain by a discrete cosine transform (DCT) or the like, each coefficient having a real value is generally quantized to represent a limited data length. In this quantization process, quantization noise is generated, resulting in lossy coding. The low frequency component has a large amplitude change and the high frequency component has a small change, and many bits are allocated to the low frequency component. In order to simply increase the signal-to-noise ratio of the recovered image, it is necessary to obtain statistical variance for each frequency component and allocate bits to maintain quantization noise uniformly regardless of frequency, but since human visual characteristics are less sensitive to quantization noise of high frequency components, The higher the frequency component, the larger the quantization step size to allow for quantization noise. To understand this, the discrete cosine transform coefficients are first divided by the weighting matrix according to the human vision, and then quantized by the quantization step size determined by the complexity of the buffer and the block. The quantization step size is uniform but does not degrade much if the quantizer output is further compressed by a variable length code (VLC), depending on statistical characteristics, but if it is followed by a variable length code (VLC) If not followed, a Lloyd-Max quantizer is required to change the quantizer step size according to the input signal amplitude, that is, to quantize it more precisely at small amplitudes with high frequency.

The quantized discrete cosine coefficients are further compressed through entropy coding using variable length codes according to their statistical characteristics. This process is a lossless coding unlike the quantization process. Entropy codes include the Huffman code, Arithmetic code, and Universal code. The quantized discrete cosine transform (DCT) coefficients are mainly variable lengths using the Huffman code. It is coded by variable length code (VLC). Huffman coding allocates bits in the original signal by assigning a few bits per code for codes with high probability of existence and many bits per code for codes with low probability of existence. The purpose of the present invention is to reduce the average transmission data rate by encoding a code having a different variable length. When the quantized discrete cosine transform coefficients are variable length coded using a Huffman code, DC coefficients and AC coefficients of the discrete cosine transform coefficients are distinguished and encoded in different methods.

Usually, the DC value of each block is highly correlated with the DC value of the neighboring block, so the difference between the DC value of the previous block is obtained and the difference value is encoded, and the DC of the first block is different from 128 which is the median of the variable range of the DC value. Obtain and encode

AC also realigns the coefficients for more efficient data compression by taking advantage of the fact that the AC coefficient near the DC coefficient is not zero in the discrete cosine transform region and that the zero is more likely to be farther from DC. At this time, the alignment is performed in one dimension through a zig-zag scan. Here, zero-run and zero-level coefficients are expressed in two dimensions (run, level). For example, a zig-zag scan is used to arrange discrete cosine transform coefficients such as 30, 2, 0, 0, -8, 0, 0, 0, 9, through run-level encoding (0). (30), (0,2), (2, -8), (3,9). In addition, when the zig-zag scanned coefficients continue to reach the end after a position, an EOB (End Of Block) code is added to indicate the end of the block. The (run, level) coded data is variable-length coded using the Huffman code table. In this case, since the probability of positive and negative values appearing in the level is almost the same, the level is taken as an absolute value and a sign is added separately. However, if the zero-run value or the level value is very large, the length of the codeword becomes too long to implement, and an escape sequence is introduced to limit the maximum code length. . In other words, if the run or level is larger than a certain value, the length of the codeword is prevented from being too long by using a fixed length such as an escape + run + level + sign.

While the bandwidth of the HDTV transmission channel is fixed, the video data is finally variable-length encoded, so the amount of data generated varies over time. Therefore, rate control (Rate control or Buffer control) is required to adjust the amount of data generated according to the given rate. The rate adjustment adjusts the amount of data generation mainly by varying the step size of the quantizer according to the buffer's fullness. That is, if the number of bits generated exceeds the reference value, the amount of data filled in the buffer increases, so the quantization step size is increased to reduce the next number of bits to occur. If data is generated below the reference value, the reverse operation is performed. Adjust to maintain. In this case, the quantization error occurs according to the quantization step size, which directly affects the image quality. Therefore, the buffer control technique is one of the most important factors that determine the image quality.

The bit rate adjustment method differs slightly depending on the video signal compression method. The target bit is set as in MPEG, the virtual buffer's fullness is calculated as the actual generation data, and the quantization step size is determined accordingly. There is a method to update the code, and the next slice is encoded by receiving a step size determined by the state of the physical buffer for each slice as used in Digicipher, CCDC, DSC-HDTV, etc. There is a way.

In MPEG, screens are basically divided into I (Intra), B (Bidirectional interpolative), and P (Predictive) to set different target bits. Calculate the target bits per slice for each mode, update to virtual buffers for each mode by the difference between the bits actually generated in one slice, and quantize the slice to be coded next according to the state of this pseudo buffer. The step size is determined and encoded. After the encoding of one frame is completed, the target bit of the next frame is calculated and continuously encoded in consideration of the amount of bits generated per frame and the remaining picture mode. Since the I mode and the P mode coexist in the screen, the quantization step size is determined according to the actual buffer fullness in slice units, unlike the rate adjustment in MPEG. That is, the data generated in one slice is filled in the buffer, and the quantization step size of the next slice is determined in proportion to the buffer state after the data has been transmitted and exited through the channel.

FIG. 2 is a diagram of a video decoder, which receives an encoded data bit stream at an input of a decoder, performs variable length decoding, and inverse quantizes to perform inverse discrete cosine transform.

Referring to FIG. 2, the buffer 32 receives an encoded bit stream transmitted from the video encoder 100 to adjust a bit rate, and the variable length decoder 34 receives a video input through the buffer 32. Variable-length encoded video data from the encoder 100 is variable-length decoded. The inverse quantizer 36 inverse quantizes to restore the state before the quantization according to the quantization level in the variable length decoder 34. Inverse quantization at the video decoder side is performed by dividing Intra DC coefficients and other coefficients (Intra AC, Inter DC, Inter AC). In the inverse quantization of the Intra DC coefficients, the DC value which has the greatest effect on the image quality among the discrete cosine transform coefficients is allocated 8-11 bits according to the precision, and 8, 4, 2, 1 corresponding to each quantization step size. Restore the DC coefficient by multiplying by When inverse quantizing Intra AC, Inter DC, and Inter AC coefficients, multiply all 64 discrete cosine transform coefficients in a block by 2, and then use the sine of the coefficient (-1 for negative, 1 for positive) only for inter-screen blocks. ), Multiply the weight matrix according to Inter / Intra, and then multiply the scale value (Quntizer_scale) of the selected quantizer according to the q_scale_type flag that distinguishes the uniform / nonuniform quantizer to obtain the inverse quantized discrete cosine transform coefficients. .

Since the inverse quantization process is a lossy process like the quantization process, some image information is lost in this process. In particular, quantization and inverse quantization of DC coefficients cause a blocky phenomenon in which a boundary between transform coding unit blocks is noticeable in a decoded output image. The inverse discrete cosine transforming unit 38 performs inverse discrete cosine transforming before the inverse quantization of the inverse quantized image by the inverse quantizer 36, and stores the image of the previous frame in the frame memory 42. do. The motion compensator 44 compensates for the motion by using the information stored in the frame memory 42 and the motion compensation information from the variable length decoder 34. The adder 40 adds the motion compensation data and the inverse discrete cosine transformed image data in the inverse discrete cosine transform unit 38 when the motion compensation is performed in the motion compensator 44.

As described above, a transform encoding such as a discrete cosine transform (DCT) in a video encoder requires a quantization process, and an inverse quantization process is required in a decoder unit of a video to restore an original image. Since the quantization and inverse quantization processes are lossy processes, some image information is lost. In particular, when quantizing and inverse quantizing the DC coefficients of the discrete cosine transform (DCT) in the conventional video encoding apparatus and video decoding apparatus, the boundary between the transform coding unit block in the output image decoded due to the quantization noise is noticeable There was a problem that the blocking phenomenon, which is a prominent phenomenon, is caused.

Therefore, in order to solve the above-mentioned problems, the present invention uses the low-pass filter as the output of the blocking phenomenon removal device only when the difference between the value passed through the low-pass filter and the pixel value of the original image is less than the threshold value. In this case, the blocking phenomenon is eliminated without blurring the fine part or the contour of the original image. In this case, the threshold value is changed depending on whether the pixel to be processed is located in a flat area or a complex area, and thus blocks in a flat area. The aim is to provide a blocking phenomenon removal device that can remove the phenomenon better and preserve the contour better in complex areas.

Another object of the present invention is to install a blocking phenomenon removal device for removing the blocking effect of the image at the output terminal of the video decoding apparatus, to remove the blocking phenomenon caused by the quantization noise of the DC coefficient to obtain a further improved image quality Its purpose is.

Blocking phenomenon removal apparatus of the present invention for achieving the above object comprises a plurality of low-pass filter for repeatedly low-pass filtering the input image; A threshold setting unit for setting a threshold according to a result of comparing a local variance value of a block with a VAR value for determining whether a pixel is located in a flat area or a complex area; And a comparison selector for outputting a low-pass filtered image only when a difference between a value passed through a plurality of low-pass filters repeatedly and a pixel value of an original image is less than a threshold value according to a threshold set by the threshold setting unit. It features.

In order to achieve the above object of the present invention, a video decoder for receiving a video encoded bit stream, variable length decoding in a variable length decoder, inverse quantization, and decoding the original video signal through inverse discrete cosine transform And a blocking phenomenon removal device for removing a blocking phenomenon of an image due to quantization noise of DC coefficients generated when the output of the variable length decoder is received and inversely quantized.

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

3 is a block diagram of an apparatus for removing blocking according to the present invention, comprising: a plurality of low pass filters 50 for repetitively low pass filtering an input image; A threshold setting unit 200 for setting a threshold value according to a result of comparing a local variance value of a block with a VAR value for determining whether a pixel is located in a flat area or a complex area; And outputting the low-pass filtered image only when a difference between a value passed through the plurality of low pass filters 50 repeatedly and a pixel value of the original image is less than a threshold value according to the threshold set by the threshold setting unit 200. Comparing selection unit 300 for. The threshold setting unit 200 includes a local variance obtaining unit 52 for obtaining a local variance for a predetermined unit block centering on a pixel currently being processed; A first comparator 54 for comparing a reference variance value VAR for determining whether the block obtained by the regional variance acquisition unit 52 is located in a flat area or a complex area according to the local variance value and the local variance. And if the local variance value is greater than the VAR value according to the result of the first comparator 54, that is, if the pixel is located in a complex area, the first threshold value Thr1 is selected and the local variance value is smaller than the VAR value. That is, when the pixel is located in a flat area, the first switch 56 selects the second threshold value Thr2. The comparison selector 300 may include: an adder 58 for adding an input image with one of Thr1 and Thr2 set by the threshold setting unit 200; A second comparator (60) for comparing the image output passing through the plurality of low-pass filters (50) with the output of the adder (58); A subtractor 62 for subtracting the input image and the threshold value set by the threshold value setting unit 200; A third comparator (64) for comparing the output of the low pass filter (50) with the output of the subtractor (62); A selection signal generator 66 for generating a selection signal by the output of the second comparator 60 and the third comparator 64 as an input; And a second switch 68 for selecting one of the input image and the output image of the low pass filter 50 according to a selection signal.

Next, the operation and effect of the blocking phenomenon removal apparatus of the present invention configured as described above will be described in detail.

The plurality of low pass filters 50 are used to repeatedly filter the input image. If the number of the low pass filters 50 is too small, the blocking phenomenon is not removed well. About three are suitable.

The threshold setting unit 200 determines a threshold value according to a result of comparing a local variance value obtained by using the input image with a reference variance value (VAR) for determining whether a block is located in a flat area or a complex area. It acts as a setting.

In the comparison selecting unit 300, only when a difference between a pixel value of the original image and a value that has passed through the plurality of low-pass filters 50 repeatedly according to the threshold value set by the threshold setting unit 200 is less than a threshold value. It functions to output low-pass images.

Looking at the action of the elements constituting the threshold setting unit 200 in detail, the regional variance acquisition unit 52 is a device for obtaining a regional variance for a block of 3 * 3 unit with respect to the pixel currently being processed, Local variance is used to determine whether the pixel being processed is located in a flat area or in a contour or complex area. The first comparator 54 compares a local variance value of the block obtained by the local variance acquisition unit 52 with a reference value VAR for determining whether a pixel is located in a flat area or a complex area according to the area variance. Do it. In this case, the VAR value is used as a reference value for determining whether the pixel is located in a flat area or a complex area according to local variance, and a value of about 30 is appropriate. The first threshold value Thr1 and the second threshold value Thr2 limit the maximum difference between the image output passing through the low pass filter 50 and the original input image. If these values are too large, the blocking phenomenon is well removed. If the original image is blurred and these values are too small, the blockage will not be removed well. The first switch 56 selects Thr1 if the local variance value is greater than the VAR value according to the result of the first comparator 54, that is, if the pixel is located in a complex area, and if the local variance value is smaller than the VAR value, When the pixel is located in a flat area, it serves to select Thr2. At this time, Thr1 should be in the relationship between Thr2 and Thr1 and Thr2 values according to the experiment are appropriate from 4 to 8.

Next, the operation of the elements constituting the comparison selecting unit 300 will be described in detail. In the adder 58, one of Thr1 or Thr2 selected by the first switch 56 in the threshold setting unit 200 is used. And adds the input image. The second comparator 60 functions to compare the image output passing through the plurality of low pass filters 50 with the output of the adder 58. The subtractor 62 subtracts the input image and the threshold selected by the first switch 56 in the threshold setting unit 200. The third comparator 64 compares the image passing through the low pass filter 50 with the output of the subtractor 62. The selection signal generator 66 performs the output of the second comparator 60 and the third. The output of the comparator 64 is used as an input to generate a selection signal. The second switch 68 serves to select one of the input image and the image passing through the low pass filter 50 according to the selection signal.

Considering the relationship between Thr1, Thr2, and VAR values and local variance, the result of the blockage removing apparatus of the present invention is represented by the following equation.

* Local variance VAR value (ie, where pixels are located in a complex area)

Threshold = Thr1

* Local variance VAR value (ie when the pixel is located in a flat area)

Threshold = Thr2

In the above formula, LPF ^N (in) is a value obtained by filtering the input image (in) N times in the low pass filter (50).

As described above, according to the present invention, the result of the low pass filtering is only output when the difference between the value that has repeatedly passed the low pass filter and the pixel value of the original image is less than the threshold value, thereby outputting the fine portion of the original image. Blocking can be eliminated without blurring the contours, since the threshold used varies depending on whether the pixel to be processed is located in a flat or complex area, so that the flat area is better removed. It is effective in preserving the contours in complex areas.

FIG. 4 is a block diagram illustrating a video decoder using a blockage removing apparatus. The video signal bit stream is received, variable length coded by the variable length encoder 82, inversely quantized, and the original video signal through inverse discrete cosine transform. In the moving picture decoder to decode by using a block-blocking phenomenon removal device 94 for removing the blocking effect of the image due to the quantization noise of the DC coefficient generated when receiving the output of the variable-length decoder 82 inverse quantization It is further provided. The blocking phenomenon removing device (94) includes: a plurality of low pass filters (50) for repeatedly low pass filtering an input image; A threshold setting unit 200 for setting a threshold value according to a result of comparing a local dispersion value of a block with a VAR value for determining whether a pixel is located in a flat area or a complex area; and the threshold value setting unit ( The comparison selection unit 300 outputs the low-pass filtered image only when a difference between a value passed through the plurality of low-pass filters 50 repeatedly and the pixel value of the original image is less than the threshold value according to the threshold value set at 200. It consists of.

Next, the operation and effects of the video decoder using the blocking phenomenon removal apparatus of the present invention configured as described above will be described in detail.

The buffer 80 receives a video encoded bit stream and adjusts a data rate. The variable length decoder 82 performs variable length encoding from a video encoder (FIG. 1) input through the buffer 80. Variable-length decoding of image data. The inverse quantizer 84 plays a role of inverse quantization in order to restore a state before being quantized according to the quantization level in the variable length decoder 82. Since the inverse quantization process is a lossy process similar to the quantization process, some image information is lost in this process.

In particular, the quantization and inverse quantization of DC coefficients cause blockiness in which a boundary between transform coding unit blocks is noticeable in a decoded output image. The inverse discrete cosine transforming unit 86 performs inverse discrete cosine transforming to the state before the inverse quantization of the inverse quantized image by the inverse quantizer 84, and the frame memory 88 stores the image of the previous frame. It works. The motion compensator 90 compensates for the motion by using the information stored in the frame memory 88 and the motion compensation information from the variable length decoder 82. Since the image reconstruction at the receiving end is estimated and compensated by the estimated motion vector value, accurate motion vector estimation is important and its compensation method is also an important problem. In the adder 92, when motion compensation is performed in the motion compensator 90, motion compensation data and the inverse discrete cosine transformed image data are added by the inverse discrete cosine transforming unit 86. The blocking phenomenon removing apparatus 94 functions to remove an image blocking phenomenon due to quantization noise of DC coefficients generated during quantization or inverse quantization.

Looking at the specific operation of the components of the blocking phenomenon removal device 94 as follows.

The plurality of low pass filters 50 are used to repeatedly filter the input image. If the number of the low pass filters 50 is too small, the blocking phenomenon is not removed well. About three are suitable. The threshold setting unit 200 sets a threshold value according to a result of comparing a local variance value of a block obtained using an input image with a VAR value that determines whether a pixel is located in a flat area or a complex area. Do it. In the comparison selecting unit 300, only when a difference between a pixel value of the original image and a value that has passed through the plurality of low-pass filters 50 repeatedly according to the threshold value set by the threshold setting unit 200 is less than a threshold value. It functions to output low-pass images.

As described above, by installing a blocking phenomenon removal device for removing the blocking effect of the image at the output terminal of the video decoder, it is possible to obtain a further improved image quality by removing the blocking phenomenon caused by the quantization noise of the DC coefficient. It works.

Claims (1)

A plurality of low pass filters 50 for low pass filtering the input image repeatedly; A threshold setting unit 200 for setting a threshold value according to a result of comparing a local variance value and a reference variance value of a block obtained by using an input image; And the low pass filter when an absolute value of a difference between a value passed through the plurality of low pass filters 50 repeatedly and the pixel value of the input image is less than a threshold value according to a threshold set by the threshold setting unit 200. A comparison selector 300 for outputting an input image and outputting the input image when an absolute value of a difference between a value passed through the plurality of low-pass filters 50 and a pixel value of the input image is greater than or equal to a threshold value The threshold setting unit 200 includes a local variance obtaining unit 52 for obtaining a local variance value for a predetermined unit block with respect to a pixel currently being processed; A first comparator (54) for comparing the regional variance value of the block obtained by the regional variance obtaining unit (52) with the reference variance value; And selecting a first threshold value when the local dispersion value is greater than the reference dispersion value according to the result of the first comparator 54, and selecting the first threshold value when the local dispersion value is smaller than the reference dispersion value, that is, the second threshold value when the pixel is located in a flat area. And a first switch 56 for selecting a value, and the comparison selecting unit 300 includes one of a first threshold value and a second threshold value set by the threshold setting unit 200 and an input image. An adder 58 for adding the signal; A second comparator (60) for comparing the images passed through the plurality of low-pass filters (50) with the output of the adder (58); A subtractor 62 for subtracting the input image and the threshold value set by the threshold value setting unit 200; A third comparator (64) for comparing the output of the low pass filter (50) with the output of the subtractor (62); A selection signal generator 66 for generating a selection signal by inputting the output of the second comparator 60 and the output of the third comparator 64; And a second switch (68) for selecting one of the input image and the output image of the low pass filter (50) according to the selection signal.