KR0124158B1 - Image decoder capable of post processing adaptively and post processing method thereof - Google Patents

Abstract

Disclosed is an apparatus and method to decompress image able to adaptively perform after-process. The apparatus performs a discrete cosine transform(DCT) and variabl length coding(VLC) to decompress the compressed image data. The apparatus comprises a variable length decompressor(10), a frequency-space domain converter(49), an after-process determiner(60), an after-processor(80) and a path controller(60). The variable length decompressor(10) decompresses the bit stream type data by reverse of the compressing sequence, and detects a quantization parameter(QP) and an moving vector(MV). The frequency-space domain converter(49) performs a reverse quantization by a coefficient of discrete cosine transform(DCT) of frequency domain according to the quantization parameter(QP), performs the reverse DCT according to the information of space domain, and performs a moving compensation according to the moving vector from the variable length decompressor(10). The after-process determiner(60) outputs a variance value of the image data from frequency-space domain converter(49), and outputs a control signal corresponding to the variance value of the image data. The after-processor(80) processes the image data from the after-process determiner(60) by the control signal from the after-process determiner(60). The path controller(60) controls path of the signals to assist the after-processor(80). Thereby, the decompress is able to be performed with the adaptive afterprocess.

Description

Image decoding apparatus and method capable of adaptive post-processing

The present invention relates to an image decoding apparatus and method capable of adaptive post-processing in an image reconstruction system for reconstructing a video signal input in a compressed form. In particular, the present invention relates to an image decoding apparatus according to a variation value of input image data. A video decoding apparatus and method suitable for performing post-processing.

As is well known, digital data is increasing in applications due to the advantage of higher accuracy than analog data. In this context, the field of processing digital signals is gradually expanding in the field of video signal processing. However, when signal processing with digital data, there are many limitations due to the huge amount of data generated.

Accordingly, various image compression techniques have been studied. Currently, a JPEG (Joint Photographic Experts Group) international standard, which is a compression method for still pictures, and an MPEG (Motion Picture Experts Group) international standard, which is a compression method for moving pictures, have been proposed. That is, in the case of JPEG, a discrete cosine transform (DCT) of an applied video signal is used to compress the video signal by removing spatial redundancy, whereas in the case of MPEG, motion compensation differential coding is used to remove temporal redundancy. The video signal is compressed by combining a technique and a discrete cosine transform technique to remove spatial redundancy.

The data compressed through such compression techniques is restored to its original form by a decoding system configured as a reverse operation of the encoding apparatus.

1 shows a typical image decoding apparatus of the related art. Data compressed through a discrete cosine transform, a quantizer, a variable length coding (VLC), a motion compensation unit, and the like, of a transmission-side encoding apparatus (not shown) are in the form of a bit stream. When applied to Variable Length Decoder (VLD) 10 Huffman decoding in the reverse process at the time of encoding, rearranges the data array in two dimensions and transfers to the dequantizer 20 do.

In this case, the VLD 10 detects a quantization parameter (QP), a motion vector (MV), and the like used during encoding applied as additional information together with an ID (Identification) code among the input data and quantizes the same. The parameter QP is transmitted to the inverse quantizer 20, and the motion vector MV is transmitted to the motion compensator 50 which will be described later.

Accordingly, the inverse quantizer 20 converts the DCT coefficient, which has been converted to a predetermined level by the quantization parameter provided from the VLD 10, to the original DCT coefficient, thereby converting the signal input from the VLD 10 into an inverse discrete cosine converter ( To 30). In addition, the inverse discrete cosine transformer 30 converts the signal output from the inverse quantizer 20 into a spatial region of the original 8x8 block in the reverse process of encoding.

On the other hand, the adder 40 directly outputs when the data output from the inverse discrete cosine converter 30 is an intra frame (that is, within a frame), but is an inter frame (that is, between frames). In contrast to the encoding process, the portion having the highest correlation among the previous frame data stored in the frame memory 51 is estimated by the motion vector (or motion vector) output from the VLD 10, as in the encoding process. When the image is output as the predicted image, the adder 40 outputs the reconstructed image signal in addition to the signal output from the inverse discrete cosine converter 30 as opposed to the time of encoding.

However, in inverse quantization during video restoration, when the quantization parameter is large, it is difficult to accurately restore the quantized DCT coefficient, which causes a blocking effect on the displayed screen. This is because the quantization of transform coefficients having a relatively high frequency among DCT transformed coefficients at the time of encoding is performed by a large step size.

Here, the region where the frequency of the DCT transformed coefficient is high is a case where the image is complicated.

Accordingly, an object of the present invention is to calculate the variance value of the input image data reconstructed into the original signal before encoding through the decoding process, and to adaptively perform post-processing of the reconstructed image data based on the calculated variance value. There is provided an image decoding apparatus capable of post-processing.

Another object of the present invention is to calculate the variance of the input image data reconstructed into the original signal before encoding through a decoding process, and to perform post-processing of the reconstructed image data adaptively based on the calculated variance value. The present invention provides a post-processing method in an image decoding apparatus.

In order to achieve the above object, the present invention provides a video decoding apparatus for reconstructing an input video signal that is compression-coded through discrete cosine transform, motion estimation compensation, and variable length coding into an original signal before encoding, wherein the video decoding apparatus includes: When the compressed and coded video signal is input in the form of a bitstream, variable length decoding is performed in the reverse process during encoding, and detecting and outputting additional information of the quantization parameter QP and the motion vector MV included in the input bitstream. Variable length decoding means; Based on the quantization parameter (QP) output from the variable length decoding means, the variable length decoded video signal in units of N × N blocks is inversely quantized into DCT coefficients in a frequency domain, and an inverse discrete cosine transform is performed into information in a spatial domain. Frequency-space domain conversion means for compensating motion based on the motion vector output from the variable length decoding means and restoring the original video signal before discrete cosine conversion; Dispersion value calculation means for calculating a dispersion value of the reconstructed N × N block unit video signal provided by the frequency-space domain conversion means; A comparator for comparing the variance value calculated by the variance value calculating means with a preset threshold value and generating a path control signal for post-processing determination of the reconstructed N × N block image signal corresponding to the comparison result; Path control means for switching the output path of the restored N × N block image signal provided by the frequency-space domain conversion means based on the generated path control signal provided from the comparator; And post-processing the first path for bypassing the restored N × N block video signal provided from the path control means to an output side and the restored N × N block video signal provided from the path control means. And a second path for providing a restored reconstructed video signal to an output side, and performing post-processing of the reconstructed N × N block video signal provided from the path control means in response to the generated path control signal provided from the comparator. Post-processing means for adaptively performing to provide a reconstructed N × N block image signal or a post-processed reconstructed N × N block image signal to an output side, wherein the post-processing means comprises; A signal for bypassing the reconstructed N × N block image signal provided through the path control means from the output side when the calculated dispersion value is smaller than a predetermined threshold value based on the path control signal from the comparator; line; And based on the path control signal from the comparator, N × N adjacent boundary pixel values of the restored N × N block provided through the path control means when the calculated dispersion value is larger than the predetermined threshold value. An adaptive post-processing image including a post-processor which provides to the output side a post-processed reconstructed N × N block image signal obtained by performing filtering that replaces the median or average value with corresponding boundary pixel values of a neighboring block. Provided is a decoding device.

In order to achieve the above object, the present invention, variable length decoding on the input video signal compression-coded by discrete cosine transform, motion estimation compensation and variable length coding, inverse quantum and inverse discrete cosine transform, this inverse discrete cosine A method of post-processing a reconstructed video signal by a video decoding apparatus which restores a compressed coded video signal to an original signal before encoding by compensating for motion of the converted signal, wherein the inverse discrete cosine transform, inverse quantization, and motion compensation are used. Restoring the compression-encoded input video signal to an original signal before encoding in units of N × N blocks; The dispersion value of the reconstructed N × N block image signal is calculated, the calculated dispersion value of the N × N block is compared with a predetermined threshold value, and the reconstructed N × N block image signal is corresponding to the comparison result. Generating a control signal for determining post-processing of the signal; In response to the generated control signal, when the calculated dispersion value of the N × N block is less than the preset threshold, bypassing the restored N × N block video signal and providing the output signal to an output side; In response to the generated control signal, when the calculated variance value of the N × N block is greater than the preset threshold, the corresponding N × N neighboring block adjacent to the boundary pixel values of the restored N × N block Calculating a median value or an average value of the boundary pixel values; And filtering the restored N × N block and providing the restored N × N block to the output side through post-processing by replacing boundary pixel values of the restored N × N block with the calculated median or average value. Provide a treatment method.

1 is a block diagram illustrating a general video decoding apparatus.

2 is a block diagram showing an image decoding apparatus according to a preferred embodiment of the present invention.

3 is a detailed block diagram of the post-processing determination unit shown in FIG.

* Explanation of symbols for main parts of the drawings

10 variable length decoder 20 dequantizer

30: Inverse Discrete Cosine Converter 40: Adder

50: motion compensation unit 60: post-processing determination unit

61 variance value calculation unit 62 comparator

70: path control means 80: post-processing unit

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

2 is a block diagram showing an image decoding apparatus according to a preferred embodiment of the present invention.

As shown in the figure, the video decoding apparatus of the present invention includes a variable length decoder 10 for decoding compressed coded image data input in the form of a bit stream in a reverse order to that in encoding, and a variable length decoder. Frequency for converting and outputting main information (actual, video signal) and additional information (for example, ID information, quantization parameter, motion vector, etc.) output from the decoder 10 into data of the original spatial domain. Calculate a variance value of the image data provided by the spatial transform unit 49 and the frequency-spatial transform unit 49, compare the calculated variance value with a predetermined reference value, and then restore the image data restored to the original signal before encoding. The frequency-space according to the post-processing judging unit 60 for generating a control signal CS for deciding whether or not to post-process and the control signal provided from the post-processing judging unit 60. Path control means 70 for controlling the path so that the restored video signal output from the inverse converter 49 is selectively outputted, and the restored video signal applied through the path control means 70 is subjected to predetermined post-processing. It consists of a post processor 80 for signal processing by the method. Here, the path control means 60 is composed of a switch (SW74), with or without post-processing control.

In particular, the frequency-space domain converter 49 performs inverse quantization on the two-dimensional array of image data output from the variable length decoder 10 by the quantization parameter QP detected by the variable length decoder 10. An inverse quantizer 20, an inverse discrete cosine transformer 30 for converting a signal output from the inverse quantizer 20 into a spatial domain, and a variable length encoder 10 when the input compressed data is inter frame data. The motion compensator 50 and the motion compensator 50 are configured to compensate for the motion by estimating a predicted image signal having a high correlation with previous frame data based on the motion vector MV. And an adder 40 for adding and outputting the predicted video signal and the signal output from the inverse discrete cosine converter 30.

Further, as shown in FIG. 3, the post-processing determination unit 60 calculates the dispersion value 61 for calculating the dispersion value of the input image data and the dispersion value obtained from the dispersion value calculation unit 61. When the control signal CS corresponding to the set threshold value (that is, the reference value) and the obtained dispersion value is greater than the threshold value, a control signal of logic '1' is generated, and the obtained dispersion value is smaller than the threshold value. In this case, a control signal of logic '0' is generated. In addition, the post-processing unit 80 may include a signal line 82 for bypassing when the dispersion value of the image data output from the adder 40 is smaller than a preset threshold value and a larger value than the preset threshold value. It consists of a post processor 71 for processing through a predetermined low pass filter (LPE).

As the filter used for such a post processor 71, a well-known median filter or a mean filter is used in the boundary part of a block, for example.

Next, a process of adaptively post-processing the video signal reconstructed to the original signal before encoding by using the video decoding apparatus of the present invention having the above-described configuration will be described.

First, the variable length decoder 10 performs an original signal (ie, quantized DCT) before variable length encoding through a process (variable length decoding) opposite to that of an encoder (not shown) with respect to an encoded input signal input in the form of a bitstream. Restore the conversion factor. In this case, the additional information corresponding to the main information, which is usually a substantial video signal, is continuously applied after the main information in a predetermined block unit.

For example, when the input video is encoded in 8x8 block units, all of the compressed video signals corresponding to the 8x8 block are applied to the additional information of the video signal corresponding to the 8x8 block. The compressed video signal corresponding to the block is applied before being applied.

That is, the variable length decoder 10 provides the quantization parameter of the detected additional information to the inverse quantizer 20, and the signal processing for the remaining additional information and the main information is processed in the same manner as in FIG. do.

In addition, the frequency-space domain converter 49 reconstructs and outputs the additional information and the main information output from the variable length encoder 10 as a signal of the spatial domain as shown in FIG. Is transmitted to the post-processing judging unit 60. In the post-processing judging unit 60, the variance value calculating unit 61 inputs an input image (i.e., image data restored to the original signal before encoding) based on Equation 1 below. The dispersion value of is calculated and provided to the comparator 62 of the next stage.

In Equation 1, N is a horizontal size, M represents a vertical size, and x (n, m) represents a (n, m) th pixel value. When the variance value is obtained by the above equation, the comparator 62 compares the preset threshold value Vthr, and provides the control signal corresponding thereto to the path control means 70, and at the same time, the frequency-space domain conversion unit ( The main information (restored image data) output from 49 is applied to the reference contact 71 of the path control means 70. The path control means 70 is controlled so that the main information (the restored image data) is switched to the second contact 73 when the control signal output from the post-processing determination unit 60, that is, the dispersion value is larger than the preset threshold value. If it is less than the threshold value, it is controlled to switch to the first contact point 72.

Next, the post processor 80 switches the signal output from the adder 40 of the frequency-space domain converter 49 when the path control means 70 is switched to the first contact point 72. Bypass through, and when the path control means 70 is switched to the second contact 73, a central filter for replacing the value of the desired time or space point with the median with the neighboring values as described above or The post-processor 81 composed of elements that apply the average value filter replacing the average value along the boundary line of the block outputs a completely reconstructed image signal.

Accordingly, in the image reconstruction method according to the present invention, when a compressed video signal is applied through a discrete cosine transform and a variable length encoding process, variable length decoding is performed in the reverse order of encoding. At this time, as described in FIG. 2, the additional information such as the quantization parameter (QP) and the motion vector (MV) among the additional information is output for the inverse quantization process and the motion compensation process through the detection process, and the inverse quantization process. And a process of calculating a variance value from the image data (restored image data) obtained through the motion compensation process and comparing it with a threshold set in advance for carrying out the present invention is performed by the post-processing determination means 60.

At this time, when the magnitude of the variance value detected in the post-processing determination process is larger than the predetermined threshold value, the post-processor 81 of the post-processing unit 80 receives the reconstructed cosine-restored signal by the frequency-space domain conversion unit 49. It is controlled by the path control means 70 to perform a post-processing process of the median filtering, or the average value filtering through. However, when the dispersion value calculated in the post-processing determination process is smaller than the predetermined threshold value, the inverse discrete cosine transformed recovery signal output from the post-processing determination unit 60 is controlled to be bypassed directly to the output side without post-processing. Therefore, through this process, a reconstructed image or a post-processed (filtered) reconstructed image is selectively generated at the output of the post-processing unit 80 in the image decoding apparatus of the present invention.

As described above, the present invention adaptively performs separate post-processing (i.e., low pass filtering) on the reconstructed video according to the magnitude of the variance value of the video signal reconstructed into the original signal before encoding in decoding the video signal. As a result, the block quality appearing on the screen due to the quantization error during display of the restored playback image can be effectively suppressed to improve the playback quality of the video receiver.

Claims (4)

An image decoding apparatus for reconstructing an input image signal compressed and encoded by discrete cosine transform, motion estimation compensation, and variable length encoding, into an original signal before encoding, the image decoding apparatus comprising :; When the compressed coded video signal is input in the form of a bitstream, the variable length decoding is performed in the reverse process during encoding, and the additional information of the quantization parameter QP and the motion vector MV included in the input bitstream is detected and output. Variable length decoding means (10); On the basis of the quantization parameter QP output from the variable length decoding means 10, the variable length decoded video signal in units of N × N blocks is inversely quantized into a DCT coefficient of a frequency domain, and inversed into information of a spatial domain. Frequency-spatial domain conversion means (49) for performing discrete cosine transform and compensating for the motion based on the motion vector output from the variable length decoding means (10) to restore the original video signal before the discrete cosine transform; Dispersion value calculating means (61) for calculating a dispersion value of the reconstructed N × N block unit video signal provided by the frequency-space domain conversion means (49); A comparator for comparing the dispersion value calculated by the dispersion value calculating means 61 with a predetermined threshold value and generating a path control signal for post-processing determination of the reconstructed N × N block image signal corresponding to the comparison result 62; On the basis of the generated path control signal provided from the comparator 62, path control means 70 for switching the output path of the restored N × N block image signal provided by the frequency-space domain conversion means 49. ); And a first path for bypassing the restored N × N block video signal provided from the path control means 70 to an output side and the restored N × N block calculation signal provided from the path control means 70. And a second path for post-processing and providing a post-processed reconstructed video signal to an output side, and provided from the path control means 70 in response to the generated path control signal provided from the comparator 62. And post-processing means 80 for adaptively performing post-processing of the reconstructed N × N block image signal to provide a reconstructed N × N block image signal or a post-processed reconstructed N × N block image signal to the output side. , The post-processing means (80); On the basis of the path control signal from the comparator 62, when the calculated dispersion value is smaller than a predetermined threshold value, the restored N × N block image signal provided through the path control means 70 to the output side. A signal line 82 for bypassing; And a boundary pixel of the restored N × N block provided through the path control means 70 when the calculated dispersion value is greater than the predetermined threshold value based on the path control signal from the comparator 62. Post-processor 81 for providing to the output side a post-processed reconstructed N × N block image signal obtained by performing filtering to replace values with median or average values of corresponding boundary pixel values of adjacent N × N neighboring blocks. Adaptive post-processing image decoding apparatus comprising a. The restored N × N block according to claim 1, wherein the path control means (70) is based on the path control signal from the comparator (62), when the calculated dispersion value is smaller than the predetermined threshold value. The image signal is transmitted to the signal line 82, and when the calculated dispersion value is greater than the predetermined threshold value, the restored N × N block image signal is transmitted to the post processor 81. An image decoding apparatus capable of adaptive post-processing. 2. The output contact according to claim 1, wherein the path control means (70) has its output contact point in accordance with the path control signal provided from the comparator (62) based on a comparison result of the calculated dispersion value and the predetermined threshold value. The image decoding apparatus capable of adaptive post-processing, characterized in that consisting of the switching post-processing control, whether or not the control switch (SW74). The variable length decoding is performed on the input video signal compressed by the discrete cosine transform, the motion estimation compensation, and the variable length encoding, followed by inverse quantization and inverse discrete cosine transform, and the motion is compensated for the inverse discrete cosine transformed signal. A method of post-processing a reconstructed video signal in a video decoding apparatus for reconstructing a coded video signal into an original signal before encoding, wherein the compressed coded input video signal is converted to N through the inverse discrete cosine transform, dequantization, and motion compensation. Restoring the original signal before encoding in × N block units; The dispersion value of the reconstructed N × N block image signal is calculated, the calculated dispersion value of the N × N block is compared with a predetermined threshold value, and the reconstructed N × N block image signal is corresponding to the comparison result. Generating a control signal for determining post-processing of the signal; In response to the generated control signal, when the calculated dispersion value of the N × N block is less than the preset threshold, bypassing the restored N × N block video signal and providing the output signal to an output side; In response to the generated control signal, the corresponding N × N neighboring block adjacent to the boundary pixel values of the restored N × N block when the calculated dispersion value of the N × N block is larger than the predetermined threshold value. Calculating a median or mean value with respect to the boundary pixel values; And filtering the restored N × N block and providing the restored N × N block to the output side through post-processing by replacing boundary pixel values of the restored N × N block with the calculated median or average value. Treatment method.