JPH0549155B2 - - Google Patents

Description

Detailed Description of the Invention (Field of Industrial Application) The present invention performs smoothing filter processing with different strengths for each moving/static block and pixel position on an image decoded by block unit processing, and further decodes the image. The present invention relates to a method and apparatus for smoothing a decoded moving image signal, characterized in that the strength of a smoothing filter is spatially changed depending on the roughness of a smoothing parameter.

(Prior Art) In order to encode image signals with high efficiency, moving image encoding/decoding systems and devices that adaptively encode mainly block by block have been widely used. Further, as a device filter for removing block distortion caused by encoding, a smoothing filter having the same characteristics for the entire screen of the decoded image signal is generally used. Alternatively, a method is known in which it is used only at block boundaries. Mitsuya et al.'s paper ``Study of image quality improvement method for block encoded images by device filter processing'', IEICE Techniques IE84-46, deals with blocks with small luminance changes where the boundaries of the blocks of the obtained decoded image become obvious. However, as shown in FIG. 9, a method is used in which filter processing is performed using eight surrounding pixels at the block boundary pixels of the decoded image or at the four corner pixels of the block. This filter processing is shown in the block diagram of FIG. The decoded image signal 51 obtained by the decoder 61 is passed through the smoothing filter 71 of the post-filter circuit 62 and the delay circuit 7.
2 and subjected to smoothing filter processing using the same characteristic coefficients, and the output image signal 53 of the delay circuit are input to the selector 73. The selector 73 receives the switching signal 5 output from the decoder 61.
According to 4, if the pixel is a pixel at the boundary of a block where the boundary of the block is obvious or a pixel in the fourth column of the block, the image signal 5 is subjected to smoothing filter processing.
2 is selected, and the output image signal 53 of the delay circuit is selected for the other pixels, and the image signal 55 is output.

(Problems to be Solved by the Invention) When a video decoding device outputs a decoded image signal as it is, distortion due to encoding becomes visible, often accompanied by visual deterioration of image quality. In particular, when encoding and decoding is performed adaptively on a block-by-block basis, the decoded image often suffers from block distortion due to encoding, as shown in a in FIG. Since such block distortion is distortion that is not familiar to ordinary TV screens, there is a problem in that it can lead to a fatal visual loss of image quality. Therefore, if a smoothing filter with the same characteristics is used for the entire screen of a decoded moving image as in the conventional general method, the resolution of the screen will decrease as a whole. In addition, in the coding and decoding of moving images, as a conventional example, Mitsuya et al.
Unlike IE84-46, block distortion does not occur in parts with little luminance change, but only moving parts are encoded and decoded, so block distortion is likely to occur in moving blocks, especially in the outline parts of moving images. There are often block boundaries. At this time, if a smoothing filter is applied only to the boundary area where block distortion occurs, the resolution of only the block boundary area will decrease, and the sharp difference in resolution between the unfiltered area and the filtered area will be noticeable. The problem is that it becomes

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a decoded moving image signal smoothing processing method and an apparatus thereof, which reduce deterioration in image quality due to encoding distortion while suppressing a decrease in overall screen resolution.

(Means for Solving the Problems) According to the first invention of the present application, there is provided a method for performing smoothing processing on a moving image signal decoded in units of blocks, wherein A method for smoothing a decoded moving image signal is obtained, which is characterized in that it discriminates between pixels in a static block area and pixels in a still block area, and switches the degree of smoothing in a smoothing filter for each pixel area in accordance with the discrimination.

Further, according to a second invention of the present application, there is provided a device that performs smoothing processing on a decoded video signal output from a video decoder based on block unit processing, which smoothes the position of a pixel within a block. a circuit for counting, a circuit for determining a pixel area to be divided by referring to an identification signal of a moving block and a stationary block in a pixel area surrounding a pixel position represented by the pixel count value of the counting circuit, and the pixel count value;
a smoothness control circuit that generates a smoothness instruction signal according to the roughness of parameters such as quantization characteristics used for decoding for each pixel region discriminated by the discrimination circuit; A decoded moving image signal smoothing processing device having a circuit for smoothing a decoded moving image signal is obtained.

(Operation) According to the present invention, the decoded image is, for example, the third
In the case of an image with block distortion such as a in the figure, amplitude discontinuities (a in Fig. 4) occur at the block boundaries. In order to visually make this decoded image a smooth image without block distortion, high frequencies are applied to pixels at the boundary of blocks with large block distortions, pixels located near these pixels, or pixels inside the blocks. By performing the smoothing process of cutting, the boundaries between blocks can be smoothed (FIG. 4, b), and an image without block distortion can be obtained, as shown in FIG. 3, b. More specifically, in order to visually smooth areas with large distortion and areas with no distortion in the decoded image, pixels at the boundaries of moving blocks (encoded blocks), pixels inside them, and still blocks (unencoded blocks) are block)
The characteristics of the smoothing filter are switched and used for each region of boundary pixels and pixels inside the boundary. In Figure 5, the strength of the filter characteristics is characteristic A<characteristic B<
Characteristic coefficients satisfying the relationship of characteristic C<characteristic D<characteristic E are prepared, and the following five areas are set as the areas in which the characteristics are switched.

A pixel area inside a block boundary pixel within a still block, and a pixel area within a still block that is three or more pixels away from the boundary with a moving block.

The pixel area surrounding the boundary in mutually adjacent static blocks (excluding the area).

A pixel area within a static block that is two pixels away from the boundary of a moving block, or at least two pixels away.

The pixel area inside the boundary pixel within a moving block.

Pixels at the boundary between a moving block and an adjacent stationary block, and pixel areas surrounding the boundary within mutually adjacent moving blocks.

When the pixel to be smoothed is a pixel within a moving block, it is determined whether the pixel is at the boundary of the block or the pixel inside it, and the pixel is divided into regions. Also,
When a pixel is in a stationary block, it is determined whether or not a moving block is adjacent by referring to the surrounding blocks. If the moving block is adjacent, it is determined whether the pixel is from the first pixel to the second pixel or the third pixel from the block boundary. It is determined whether the pixel is an inner pixel that is further away from the pixel, and the pixel is divided into an area with the area and an area with the area. Furthermore, it is determined whether a pixel is a pixel at the boundary of mutually adjacent still blocks or a pixel inside the boundary, and the pixel is divided into regions. If the pixel divided in this way is in the area, select filter characteristic A, if in the area, select filter characteristic B, if in the area, select filter characteristic C, and if in the area, select filter characteristic D. When the area is selected, filter characteristic E is selected and used. That is, by switching the strength of the filter characteristics as shown in <<<< for the five pixel regions, the boundaries of blocks including contour parts can be smoothly extended.

In the above explanation, the pixel was divided into ~ regions, but by combining the five regions, the following (a)
You may choose something like ~(p). FIG. 6 shows area divisions (a) to (l).

(a) Divide into two regions, + and ++.

(b) Divide into two regions, ++ and +.

(c) Divide into two regions, +++.

(d) Divide into two regions, + and ++.

(e) Divide into two areas: ++ and +.

(f) Divide into two regions, +++.

(g) Divide into three areas: , , ++.

(h) Divide into three areas: +, , +.

(i) Divide into three areas: ++, .

(j) Divide into three areas: +, , +.

(k) Divide into three areas: , +, +.

(l) Divide into four areas: +, , .

Divide into four areas: (m), +, .

(n) Divide into four areas: +, , , .

(o) Divide into four areas: , , +.

(p) Divide into four areas: , , +.

Furthermore, as shown in FIG. 7, the strength of the filter characteristics is changed depending on the coarseness of the decoding parameters. For example, when the quantization characteristic is coarse, the distortion becomes large, so filter characteristic E, which has a strong blur in the pixel region shown in FIG. 5, is used. On the other hand, when the quantization characteristics are rather fine, block distortion does not appear conspicuously, so the filter characteristic A with the weakest blur is used in the pixel region of FIG. 5, and the filter characteristic C with the weakest blur is used in the pixel region of FIG. Filter characteristic C is used for the pixel region of and . In this way, by changing the characteristics of the smoothing filter over time according to the roughness of the decoding parameters, when the encoding distortion is small, the reduction in the resolution of the decoded image is suppressed and the encoding distortion is reduced. It has the characteristic that the image quality can be improved when the value is large.

(Embodiment) Referring to FIG. 1 and FIG. 2, details of an embodiment in which the present invention is applied to a moving picture signal decoding device that decodes a signal obtained by encoding a moving picture signal in units of blocks on the receiving side. Explain. Note that in this embodiment, each signal is a digital signal. Also, the transmitted signal is
Assume that TV line scanning has been converted to line scanning in blocks.

1 and 2 respectively show the configurations of the first and second embodiments of the device according to the second invention of the present application, and below, together with the embodiments of the second invention, the configuration of the first embodiment of the apparatus according to the second invention of the present application will be described. First and second embodiments of the invention will also be described.

First, the embodiment shown in FIG. 1 will be described.

In FIG. 1, a received compressed encoded signal 1 is once stored in a buffer memory 101, and a compressed encoded signal 2 is outputted in accordance with the timing of decoding. The compressed encoded signal 2 is decompressed and decoded in the decoder 102, and a decoded image signal 3 is output. Furthermore, the block synchronization signal 6, dynamic block identification signal 8, and decoding parameters 10 used for the decoding process are output. The block synchronization signal 6 is input to a pixel address counter 103, which outputs a pixel count value 7 indicating the position of the pixel within the block. Pixel count value 7 and dynamic block identification signal 8
is input to the pixel area determination circuit 104. The pixel area discrimination circuit 104 uses the pixel count value 7 and the dynamic block identification signal 8 to output an identification signal 9 representing five pixel areas. First, when the moving/static block identification signal 8 indicates a moving block area, it is determined whether the pixel to be subjected to filter processing is a boundary pixel or a pixel inside thereof based on the pixel count value 7, and as shown in FIG. Divide the area into two areas. In addition, when representing a stationary block area, it is determined whether the moving blocks are adjacent by referring to the surrounding moving/static block identification signals, and if the moving blocks are adjacent, one pixel from the block boundary is determined. It is determined whether the pixel is the second pixel or the inner pixel that is three or more pixels away, based on the pixel count value 7, and the pixel is divided into an area with the area and an area with the area. Furthermore, it is determined whether a pixel is a pixel at the boundary of mutually adjacent still blocks or a pixel inside thereof, based on the pixel count value 7, and the pixel is divided into regions. The identification signal 9 representing the five pixel areas divided in this way is input to the filter characteristic control circuit 105, and when the decoding parameter 10 is coarse quantization, the filter characteristic E with strong blur is applied to the area in the moving block. , and select a slightly stronger filter characteristic C for the area within the stationary block. Furthermore, when the decoding parameter 10 is fine quantization, filter characteristic B with weak blur is selected. That is, a filter characteristic control signal 11 is outputted for selecting a filter characteristic suitable for the roughness of the decoding parameter 10 as shown in FIG. 7 for the pixel regions .about. shown in FIG. 5. Similar to the decoded image signal 3, this filter characteristic control signal 11 is output in line scanning order in block units. The decoded image signal 3 and the filter characteristic control signal 11 are converted from block-by-block line scanning to TV line scanning in scan converters 106 and 107, respectively.
The decoded image signal 4 and the filter characteristic control signal 12 are output pixel by pixel in the order of line scanning. The image signal 4 is input to a smoothing filter 108, which performs smoothing processing by switching the strength of the filter characteristic (characteristic coefficient) for each pixel area according to the filter characteristic control signal 12, and outputs the image signal 5. The smoothing filter 108 in FIG. 1 performs smoothing processing by multiplying the surrounding pixels and the center pixel by a characteristic coefficient, and therefore requires a line memory of at least three lines.

Next, the embodiment shown in FIG. 2 will be described.

In FIG. 2, the received compressed encoded signal 21
is stored once in buffer memory 201, and outputs compressed encoded signal 22 in accordance with the timing of decoding. The compressed encoded signal 2 is decompressed and decoded in a decoder 202, and a decoded image signal 23 is output. Furthermore, it outputs a line synchronization signal 26, a dynamic block identification signal 28 used for decoding processing, and decoding parameters 30. Here, the dynamic block identification signal 28 and the decoding parameter 30
is output by line scanning in block units during decoding. The line synchronization signal 26 is input to the pixel address counter 203, which outputs a pixel count value 27 indicating the position of the pixel within N lines for the block size N×N. Dynamic block identification signal 2
8 and decoding parameters 30 are scan converter 204
The signal is converted into TV line scanning in the TV line scanning order, and a motion/static identification signal 31 and a decoding parameter 32 are output pixel by pixel in the TV line scanning order. The pixel count value 27 and the motion/static identification signal 31 are input to the pixel area determination circuit 205 . The pixel area discrimination circuit 205 uses the pixel count value 27 and the motion/static identification signal 31 to output an identification signal 29 representing five pixel areas. First, when the motion/static identification signal 31 indicates a motion block area, it is determined whether the pixel to be subjected to filter processing is a pixel at the block boundary or a pixel inside the block based on the pixel count value 27, as shown in FIG. Divide into two areas. Furthermore, when the motion/static identification signal 31 represents a stationary block area, it is determined whether or not the pixels of the dynamic area are adjacent by referring to the surrounding dynamic/static identification signals, and if the moving areas are adjacent, Based on the pixel count value 27, it is determined whether the pixel is the first pixel, the second pixel, or an inner pixel separated from the block boundary by three or more pixels, and is divided into an area with the area and an area with the area. Furthermore, it is determined whether the pixel is a pixel at the boundary of mutually adjacent still block regions or a pixel inside the boundary, based on the pixel count value 27, and the pixel is divided into regions. The identification signal 29 representing the five pixel areas divided in this way is input to the filter characteristic control circuit 206, and the decoding parameter 32 is input to the filter characteristic control circuit 206.
When quantization is coarse, a filter characteristic E with a strong blur is selected for a region of a moving block, and a filter characteristic C of a slightly strong blur is selected for a region of a stationary block. In addition, when the decoding parameter 32 is fine quantization, the filter characteristic B has weak blur.
Select. That is, a filter characteristic control signal 33 for selecting a filter characteristic suitable for the coarseness of the decoding parameter 32 as shown in FIG. 7 is output for each pixel for the pixel regions .about. shown in FIG. 5. The decoded image signal 23 is converted from block unit line scanning to TV line scanning in a scan converter 207, and the decoded image signal 23 is converted into a TV line scanning order.
Make it 4. The decoded image signal 24 is passed through the smoothing filter 2
08, the strength of the filter characteristic is switched for each pixel region according to the filter characteristic control signal 33, smoothing processing is performed, and an image signal 25 is output. In the smoothing filter 208 in FIG.
Smoothing processing is performed by multiplying the surrounding pixels and the center pixel by a characteristic coefficient, and therefore a line memory of at least three lines is required.

In the embodiment of FIGS. 1 and 2, the smoothing filter circuit 108, 208 is connected to the scan converter 107, 2.
Although it is placed after 07, the scan converter 10
It may be placed before 7,207. In this case, scan converters 106, 204 are not needed. Also,
Smoothing filters 108 and 208 require two or more block lines of memory.

In the above description, a case has been described in which the pixel area is divided into five parts, but it may also be divided into two pixel areas as shown in (a) to (f) shown in FIG. 6. Also,
It may be divided into three pixel areas as shown in (g) to (k). Furthermore, it may be divided into four pixel areas as shown in (l) to (p).

(Effects of the Invention) According to the present invention, the smoothing filter makes the distortion at the block boundary less noticeable, and the encoding distortion can be significantly reduced, so that a large visual improvement in image quality can be obtained. In addition, in the present invention, smoothing processing is performed by changing the characteristics of the smoothing filter according to pixel regions with different encoding distortions and the degree of encoding distortion, so it is possible to reduce the decrease in resolution of the entire decoded image,
A decoded image that is visually easy to see is obtained.

[Brief explanation of drawings]

Figures 1 and 2 are block diagrams showing embodiments of the invention, Figures 3a and b are diagrams for explaining the effects of the invention, and Figures 4a and b are diagrams for explaining the effects of the invention. FIGS. 5 and 6 are diagrams for explaining the method of dividing a pixel area in the present invention, FIG. 7 is a diagram for explaining the filter processing control method of the present invention, and FIG. FIG. 9, a block diagram showing a conventional example, is a diagram for explaining the conventional method. 1... Received compressed encoded signal, 2... Compressed encoded signal, 3... Decoded image signal, 4... Decoded image signal, 5... Smoothing filter output image signal, 6...
Block synchronization signal, 7...Pixel count value, 8...
...Dynamic block identification signal, 9...Pixel area identification signal, 10...Decoding parameter, 11...Filter characteristic control signal, 12...Filter characteristic control signal, 101...Buffer memory, 102...Decoder, 103 ...Pixel address counter, 104...
...Pixel area discrimination circuit, 105...Filter characteristic control circuit, 106...Scan converter, 107...Scan converter, 108...Smoothing filter, 21...Received compression encoded signal, 22...Compression encoding signal, 2
3...Decoded image signal, 24...Decoded image signal, 2
5... Smoothing filter output image signal, 26... Line synchronization signal, 27... Pixel count value, 28...
...Moving and static block identification signal, 29...Pixel area identification signal, 30...Decoding parameter, 31...Motion and static identification signal, 32...Decoding parameter, 33
... Filter characteristic control signal, 201 ... Buffer memory, 202 ... Decoder, 203 ... Pixel address counter, 204 ... Scan converter, 205 ...
... Pixel area discrimination circuit, 206 ... Filter characteristic control circuit, 207 ... Scan converter, 208 ... Smoothing filter, 50 ... Compression encoded signal, 51 ...
Decoded image signal, 52... Smoothing filter output image signal, 53... Delay circuit output signal, 54... Switching signal, 55... Output image signal, 61... Decoder, 62... Post filter circuit, 71...Smoothing filter, 72...Delay circuit, 73...Selector.

Claims (1)

[Claims] 1. In a method of performing smoothing processing on a moving image signal decoded in block units, the moving image signal is distinguished between pixels in a moving block area and pixels in a still block area, and A decoded video signal smoothing processing method characterized by switching the degree of smoothing in a smoothing filter for each pixel region according to determination. 2. In a device that performs smoothing processing on a decoded video image signal output from a video decoder based on block unit processing, there is a circuit that counts the position of pixels within a block, and a pixel count value of this counting circuit. A circuit that identifies the pixel area to be divided by referring to the identification signal of moving blocks and static blocks in the pixel area surrounding the pixel position represented by the pixel position and the corresponding pixel count value, and decoding for each pixel area determined by this determination circuit. a smoothness control circuit that generates a smoothness instruction signal according to the roughness of parameters such as quantization characteristics used in the process; and a circuit that performs smoothing processing of the decoded video signal in accordance with the smoothness instruction signal. A decoded video signal smoothing processing device comprising: