JP2831139B2 - Image signal decoding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image signal decoding apparatus for decoding an image signal transmitted or recorded after being highly compressed and encoded, and more particularly to a block distortion removing process in such an apparatus. .

[0002]

2. Description of the Related Art Generally, an image signal captured by a solid-state imaging device represented by a CCD (Charge Coupled Device) is recorded as digital data on a recording device such as a memory card, a magnetic disk, or a magnetic tape. , The amount of data is enormous. Therefore, in order to record such data in a limited recording capacity range, it is necessary to perform some highly efficient compression on the obtained image signal data. As a high-efficiency image data compression method, an encoding method using orthogonal transform encoding is widely known. An example of this method will be described with reference to FIG.

First, when image data (f) is input from a solid-state image pickup device or the like (101), the image data (f) is divided into blocks of a predetermined size to obtain a value (f _b ).
Is obtained (102), and a two-dimensional DCT (discrete cosine transform) is performed as an orthogonal transform for each of the divided blocks to convert it into a value (F) (103). Next, linear quantization corresponding to each frequency component is performed (104), Huffman coding is performed on the quantized value (F _Q ) as variable length coding (105), and the result is compressed data ( Transmitted or recorded as C). At this time, the quantization width of the linear quantization is determined by preparing a quantization matrix representing relative quantization characteristics in consideration of visual characteristics for each frequency component, and multiplying the quantization matrix by a constant to obtain a quantization width. Has been determined.

On the other hand, when image data is generated from the compressed data, a quantized value (F _Q ) of a transform coefficient is obtained by decoding (decoding) the variable length code (C) (106).
However, it is impossible to obtain a true value (F) before quantization from this value, and the result obtained by inverse quantization is a value (F ′) including an error (107). Therefore, IDCT (inverse discrete cosine transform) is performed on this value (F ′) (10
8), the resulting value (f _b ′) is deblocked (1
09) The obtained image data (f ′) also includes an error. Therefore, the image quality of the (110) reproduced image (f ′) reproduced and output by the image reproducing device or the like deteriorates.
That is, the error of the value (F ′) obtained as a result of the inverse quantization is a so-called quantization error, which causes deterioration of the image quality of the reproduced image (f ′).

The above operation will be specifically described with reference to FIG. First, as shown in FIG. 10A, one frame of image data is divided into blocks of a predetermined size (for example, 8 blocks).
), And two-dimensional DCT is performed for each of the divided blocks as an orthogonal transform, and the blocks are sequentially stored on an 8 × 8 matrix.

[0006] When viewed on a two-dimensional plane, the image data has a spatial frequency which is frequency information based on the distribution of density information. Therefore, by performing the DCT, the image data is converted into a DC component DC and an AC component AC as shown in FIG. 10B, and the origin position ((0, 0) DC component D at position)
The data indicating the value of C, the data indicating the maximum frequency value of the AC component AC in the horizontal axis direction at the (0, 7) position, and the AC component component in the vertical axis direction at the (7, 0) position. The data indicating the maximum frequency value of the AC is stored at the (7, 7) position, and the data indicating the maximum frequency value of the AC component AC in the oblique direction is stored at the position (7, 7). At the intermediate position, the frequency data in the direction associated with each coordinate position is stored in such a manner that the frequency data having higher frequency than the origin side appears.

Next, the data stored at each coordinate position in the matrix is divided by the quantization width for each frequency component to perform linear quantization according to each frequency component. Performs Huffman coding as variable length coding. At this time, regarding the DC component DC, the difference value between the DC component and the DC component of the neighboring block is Huffman-coded.

As for the AC component AC, a scan from a low frequency component to a high frequency component called a zigzag scan is performed, and the continuous number of invalid (value is “0”) components (zero number of runs) and the succeeding number. Valid component value 2
Dimensional Huffman encoding is performed to obtain encoded data.

In this method, the compression ratio is generally controlled by changing the quantization width of the quantization. The higher the compression ratio, the larger the quantization width.
Therefore, the quantization error becomes large, and the deterioration of the image quality of the reproduced image becomes conspicuous.

The quantization error of the transform coefficient tends to appear as a so-called block distortion in which a discontinuity occurs at a block boundary portion in a reproduced image, and this block distortion is visually conspicuous. Is good,
The subjective impression gets worse.

Therefore, a method has been devised in which a low-pass (low-pass) filter is applied to the image reproduced by the decoder as distortion removal processing. This post-filter can remove distortion relatively satisfactorily, but when the image contains edges and the like, they are blurred, and conversely, the degree of low-pass is reduced to reduce blur. Then, there is a problem that the block distortion cannot be completely removed.

In order to solve this problem, the presence or absence of an edge in an image and block distortion are detected, and whether or not to apply a filter is switched according to the result. Methods are also known.

[0013]

However, as described above, in a distortion removal method in which a filter is applied only to a portion where distortion exists, an image having many fine structures or an image in which edges exist particularly on block boundaries are used. In the case of an image, the edge is filtered without being able to distinguish the edge from the distortion, so that there is a disadvantage that the image is still blurred. Moreover,
Depending on the method of detecting block distortion in an image, there have been problems such as a long processing time and difficulty in removing distortion.

The present invention has been made in view of the above points, and an image signal decoding apparatus capable of removing block distortion without causing blurring or the like in an image by using a simple circuit. It is intended to provide.

[0015]

In order to achieve the above object, an image signal decoding apparatus according to the present invention divides image data into blocks, and performs an orthogonal transform for each of the divided blocks. Variable-length code decoding means for quantizing the transformed output and thereafter decoding the compressed image data by performing variable-length encoding on the quantized output;
Inverse quantization means for obtaining an orthogonal transform coefficient by inversely quantizing the decoded output from the variable length code decoding means,
A first inverse orthogonal transform unit for performing an inverse orthogonal transform on the orthogonal transform coefficient from the inverse quantization unit, a distortion removing unit that performs a distortion removal process on a transform output from the first inverse orthogonal transform unit, Orthogonal transformation means for orthogonally transforming the distortion removal processing output from the distortion removal means to obtain an orthogonal transformation coefficient; and an orthogonal transformation coefficient before and after the distortion removal processing for each block from both outputs of the inverse quantization means and the orthogonal transformation means. The amount of change is calculated, and the calculated amount of change in the coefficient is compared with the maximum value of the quantization error of the coefficient, and the amount of change is the maximum value of the value of the quantization error of each coefficient. When the change amount is larger than the maximum value of the possible value of the quantization error of each coefficient, the clipping unit that corrects and outputs the orthogonal transformation coefficient from the orthogonal transformation unit, and Modified orthogonal transformation And a second inverse orthogonal transform means for obtaining a decoded reproduction signal by inverse orthogonal transform coefficients.

[0016]

In the image signal decoding apparatus according to the present invention, the orthogonal transformation means orthogonally transforms the distortion removal processing output to obtain an orthogonal transformation coefficient, and the clipping means produces the first orthogonal transformation coefficient.
The amount of change from the orthogonal transform coefficient which is an input to the inverse orthogonal transform means is obtained, and the amount of change is compared with the maximum value of a possible quantization error of the coefficient. When the quantization error is larger than the maximum value of the possible values of the quantization error, the orthogonal transformation coefficient from the orthogonal transformation means is corrected and output so that the change amount is equal to or less than the maximum value of the possible value of the quantization error of each coefficient. The modified orthogonal transform coefficient is inversely orthogonally transformed again by the second inverse orthogonal transform means to obtain a decoded reproduced signal.

That is, orthogonal transformation is again performed on the image after the distortion removal processing, and how much the transform coefficient of each block has changed before and after the distortion removal processing is obtained. When the value is larger than the maximum value that can be taken, it is understood that components existing in the image signal have been lost by the distortion removal processing, so the amount of change in the coefficient is equal to or less than the maximum value that the quantization error of each coefficient can take The lost information is recovered by performing clipping to correct the information.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Before describing the embodiments of the present invention, the concept of the present invention is first described with reference to FIG.
This will be described with reference to FIG.

First, the recorded compressed image data is
The data is read by the variable length code decoding circuit 10. The output of the variable length code decoding circuit 10 is returned to an orthogonal transform coefficient by an inverse quantization circuit 11. The orthogonal transform coefficients are converted into real space data by the inverse orthogonal transform circuit 12, and then the distortion removal circuit 1
At 3, distortion removal processing is performed. The orthogonal transform circuit 14 converts the result into orthogonal transform coefficients again. Then, the clipping circuit 15 outputs the output of the orthogonal transform circuit 14 and the coefficient change of the orthogonal transform coefficients of the same block obtained by the inverse quantization circuit 11. When the amount of change is greater than the maximum value of the quantization error of each coefficient, the amount of change is equal to or less than the maximum value of the quantization error of each coefficient. Thus, the transform coefficient after the distortion removal processing is corrected. The corrected coefficients are again converted into real space data by the inverse orthogonal transform circuit 16, and then displayed and output on the output device as a reproduced image.

In general, the degree of conspicuousness of block distortion changes depending on the spatial frequency of a nearby image. That is,
When block distortion occurs in a portion having components up to a high frequency having a fine structure, the block distortion is not so noticeable. Conversely, when block distortion occurs in a portion having only low spatial frequency components that change relatively slowly, the block distortion becomes more conspicuous.

On the other hand, block distortion is caused by discontinuity at a block boundary, and therefore has components up to a very high spatial frequency. Therefore, by removing spatial frequency components higher than the spatial frequency of the image near the distortion, block distortion can be made inconspicuous. However, where the image contains high frequency components,
Since even the high frequency components of the image data are lost together, the image is blurred.

The quantization error of the transform coefficient, which is the cause of the block distortion, is caused by dividing the transform coefficient into quantization widths and replacing them with respective representative values. In other words, since the change caused by the process of removing the distortion should change so that the quantization error of the transform coefficient decreases, it is considered that the value of the transform coefficient before the distortion removal process was included. It should not change beyond the maximum quantization distortion.

This situation will be described with reference to the drawings. In a general quantizer, as shown in FIG. 2, an input value from i ₀ to i ₁ which is larger by a quantization width is represented by an average value q Outputs ₀ as the representative value. Similarly, and converts the input value from i ₁ to i ₂ input values from q _1, i ₂ to i ₃ to q _2. Therefore, the true value cannot be obtained from the value once quantized. For example, given a quantized value q ₁ , its true values are i ₁ and i
_It can be seen that the value was between ₂ , and the error contained a maximum of half the quantization width with respect to the true value. Conversely, when we perform the processing as to reduce the quantization distortion, the values of q ₁ is gradually varied between i ₁ and i _2, the case that beyond this, the deterioration of the image information Is thought to be happening.

Therefore, in the present invention, the image after the distortion removal processing is again subjected to the orthogonal transformation, and it is determined how much the transform coefficient of each block has changed before and after the distortion removal processing. When the quantization error is larger than the maximum value that can be taken, it is known that components existing in the image signal have been lost due to the distortion removal processing. Lost information is restored by performing clipping to correct the information to be less than the maximum value that can be obtained. Hereinafter, embodiments of the present invention will be described with reference to the drawings.

Here, it is assumed that the recorded or transmitted image data has been subjected to high compression encoding as follows. That is, in the encoding apparatus, a discrete cosine transform (DCT) is used for the orthogonal transform, and the orthogonal transform is performed by a quantizer. When D ≧ 0: Q = int ｛(D + Q _W / 2) / Q _W ＤD <0 At the time of: Q = int {(D−Q _W / 2) / Q _W } (1) Where D is the input DCT coefficient,
Q represents an output, Q _W represents a quantization width, and int represents truncation below a decimal point. Further, a Huffman code is used as the variable length code.

FIG. 3 is a block diagram of an image signal decoding apparatus according to the first embodiment of the present invention, which decodes the result of the encoding. That is, the signal passed through the Huffman decoding circuit 10A is a value represented by Q in the equation (1). Subsequently, the signal that has passed through the inverse quantization circuit 11 is set to a value obtained by multiplying Q by Q _W as represented by Q _i = Q × Q _W (2). This Q _i is a value including a quantization error of the DCT coefficient.

This signal is divided into two, one of which is once recorded on a block-by-block basis in the memory 19 connected to the clipping circuit 15, and the other is sent to the IDCT (inverse discrete cosine transform) circuit 12A.

The IDCT circuit 12A uses the IDCT circuit 12A for each block.
The T-signal is sequentially written into the frame memory 18 as a reproduced signal after the data of all blocks is written, and then sent to the distortion removing circuit 13 and subjected to a low-pass filter.
The result is divided into exactly the same blocks as before, and D
After being subjected to DCT by the CT circuit 14A, the signal is sent to the clipping circuit 15. Then, it is compared with the transform coefficient Q _i stored in the memory 19 before performing the distortion removal processing.

By this comparison, it is determined how much each coefficient changes before and after the distortion removal processing. When the coefficient Q ′ after the distortion removal processing is Q ′ ≧ 0: Q ′ ≧ (Q _i + Q _W / 2) or Q ′ <(Q _i −Q _W / 2) When Q ′ <0: When Q ′> (Q _i + Q _W / 2) or Q ′ ≦ (Q _i −Q _W / 2), the clipping circuit Reference numeral 15 outputs the coefficients before the distortion removal processing, and outputs the coefficients after the distortion removal processing when they are not satisfied. This output is subjected to IDCT by an IDCT circuit 16A to obtain a reproduced image.
Display output.

At this time, the IDCT circuit 12A and the IDCT circuit 16A may have a common configuration. The distortion removal processing in the distortion removal circuit 13 may be performed by a convolution filter in a real space, may be performed by spatial frequency filtering such as a Fourier transform plane, or may be any other processing that reduces distortion. Any method can be used. Next, an image signal decoding apparatus according to a second embodiment of the present invention will be described with reference to FIG.

The signal that has passed through the Huffman decoding circuit 10A and the inverse quantization circuit 11 is divided into two, as in the first embodiment shown in FIG.
The data is recorded in a memory (not shown) incorporated therein, and the other is sent to the IDCT circuit 12. The signals subjected to IDCT for each block by the IDCT circuit 12 are sequentially sent to a distortion removal circuit 13, where the data for all blocks is converted into a reproduction signal, and then subjected to a low-pass filter. The output of the distortion removal circuit 13 is supplied to the switching circuit 20 by a DCT.
It is selectively switched and supplied to the circuit 14A or the output device 17. When this signal is sent to the DCT circuit 14A, it is converted into a DCT coefficient as in the first embodiment, and then clipped by the clipping circuit 15.

The clipping circuit 15 sends out a signal for incrementing the value of a repetition counter (not shown) incorporated in the repetition determination circuit 21. The result of the clipping is sent to the IDCT circuit 12. The clipping result from the clipping circuit 15 is subjected to IDCT again in the IDCT circuit 12 and sent to the distortion removal circuit 13.

The repetition determination circuit 21 uses the value of the repetition number counter to determine whether or not to perform the second and subsequent distortion removal processing and clipping. When the distortion removal processing is performed, an instruction is given to the distortion removal circuit 13 so that the filtering is performed in the same manner as the previous time. At the same time, when it is determined that clipping is not to be performed, the connection destination of the switching circuit 20 is switched to the output device 17.

By doing so, it is possible to repeatedly execute the distortion removal processing and clipping until the connection destination of the switching circuit 20 is switched to the output device 17, and when conceptually expressed, it is shown in FIG. It is possible to perform a loop process that can escape from both the distortion removal process and the clipping.

In this method, it is possible to use a method in which a filter is applied in a plurality of times in the distortion removal processing. For example, a filter having a gradual frequency cutoff characteristic is applied in a plurality of times. With such a multi-stage filter, a method of obtaining a sharp frequency cutoff characteristic can be used. At this time, if the number of filter stages is known,
If you don't need to clip every time in the loop,
The clipping may be omitted. Also,
The present invention is based on the number of times distortion is removed in a loop.
It is preferable to have a function of changing the distortion removal characteristics.

For example, when a convolution filter is used as the distortion removal filter, a complicated frequency cutoff characteristic can be obtained by increasing the kernel size, but on the contrary, the amount of calculation increases. However, the circuit becomes large. Therefore, in the present embodiment, it is possible to use a method of realizing a filter having desired characteristics by combining a plurality of small-sized filters.

Further, according to this method, it is possible to gradually reduce the possibility that the distortion reduced by filtering appears again due to clipping. In other words, if only one transform coefficient in a block is changed by clipping, the effect appears on the entire block, and block distortion may appear again. However, in general, 1
Since the degree distortion removal processing is performed, the amount of distortion is smaller than that of the data that first enters the distortion removal circuit 13. Therefore, the degree of low-pass may be weaker in the second distortion removal processing than in the first distortion removal processing. Similarly, after the third time, the degree of the low-pass is further reduced, and by repeating an appropriate number of times, the signal is output when the need for the distortion removal processing is eliminated. Next, a third embodiment of the present invention relating to a method for adaptively performing the distortion removal processing for each block using this method will be described.

In the third embodiment, in order to remove block distortion from the result obtained by decoding, the transform coefficient of each block is compared with a predetermined value, and the result is calculated as follows. , A significant coefficient and a non-significant coefficient, and the characteristic of distortion removal is changed for each block based on a frequency band corresponding to the significant coefficient.

FIG. 6 is a block diagram of the third embodiment. That is, the variable length code of the compressed image data is decoded by the Huffman decoding circuit 10A, and the inverse quantization circuit 1
Since the result of inverse quantization at 1 is a transform coefficient for each block, it corresponds to a spatial frequency component in the block. This data is divided into two, one of which is an IDCT circuit 1
2 to obtain an image signal in the real space. Further, the distortion removal characteristic of the image signal in the real space is determined using the other.

The method for determining this characteristic is as follows:
In step 2, the absolute value of the spatial frequency component is compared with a threshold value for each block, and a coefficient larger than the threshold value is defined as a significant coefficient, and the distortion removal characteristic determination circuit 23 saves the band of the significant coefficient. To determine. Then, the distortion removing circuit 13 performs a distortion removing process of the determined characteristics.
Other operations of the DCT circuit 14A and the clipping circuit 15 are the same as those of the second embodiment.

With such a configuration, strong low-pass filtering for averaging over a wide range is performed on a block having only low spatial frequency components, and conversely, on a block including even relatively high spatial frequency components. By performing weak low-pass filtering that does not blur much, low-pass filtering that does not blur the structure in the block can be realized. Hereinafter, this state will be specifically described.

Now, the orthogonal transformation coefficient of the block of interest given to the coefficient determination circuit 22 is calculated as shown in FIG.
X8 pixel DCT coefficients, and the absolute value of each coefficient is set to a threshold th
Compared with (th = 10), the coefficient of the portion indicated by hatching in FIG. 7B is determined to be a significant coefficient having an absolute value of “10” or more. In this case, it can be seen that this block includes a structure that can be almost completely expressed by information up to half the maximum frequency (1/2 of the sampling frequency) in the horizontal and vertical directions. Therefore, the filter characteristic for this block should be such that it cuts frequencies higher than half the maximum frequency in both the horizontal and vertical directions. For simplicity, this situation will be considered in one dimension only in the horizontal direction in the above example.

As described above, if it is found that the block of interest is composed of up to half the frequency component of the highest frequency (f _max ) from the one corresponding to the highest frequency among the significant coefficients, then (f _max / 2), so that components higher than this band may be cut. That is, the band may be cut with the passband indicated by hatching in FIG.

This can be expressed by the following equation: H = F × G (3) Here, equation (3) represents filtering in the frequency domain, where F, G, and H are data,
The filter and the processing result are coefficients on the Fourier plane, and the G component may be as shown in FIG.

If this filtering is realized by convolution in the real space and the convolved coefficient is changed, it is expressed as h = f * g (4). Note that, in equation (4), f,
g and h are the inverse Fourier transform results of F, G and H, respectively, and represent the case where the filtering of equation (3) is processed in real space. As in this equation (4), the filtering is a convolution of the data f and g.
Since the kernel size of g is finite, it is impossible to obtain a sharp cutoff characteristic as shown in FIG. And cutoff characteristics.

When the characteristic of the first distortion elimination filter is determined for each block in the distortion elimination characteristic determination circuit 23 as described above, it is stored for the time being, and filtering is performed on each block by the characteristic. . Subsequently, clipping is performed to make the low-pass characteristics of each block weaker in the second filtering than in the first filtering. In other words, the filtering is performed so that the band is cut in a range wider than the hatched portion in FIG. Similarly, in the third filtering, a wider range is set as a passband, and this is repeated, and when the filter characteristics acting on all blocks have become sufficiently weak, the loop is exited. At this time, filtering is not required earlier in a block whose filter characteristics are not strong as compared with other blocks. In such a case, subsequent filtering is omitted for the block.

To determine whether a conversion coefficient is a significant coefficient, instead of comparing the absolute value of each coefficient with a threshold value as in the third embodiment, the value of each coefficient is zero. It may be determined based on whether or not. Moreover,
If there are consecutive blocks that have only very low or no significant coefficients at all, consider them as a macroblock and apply a strong low-pass to this macroblock over a wide range. It may be something that you have.

The present invention is not limited to the block size, the type of orthogonal transform, the type of variable length coding, and the like used in the above-described embodiments. Further, the distortion removal filter may be applied separately in the horizontal direction and the vertical direction, or a two-dimensional filter may be applied at one time. Is also good.

[0049]

As described in detail above, according to the present invention,
Since block distortion can be removed at high speed without blurring in the image, and the circuit is simple,
It is possible to provide an image signal decoding device that can be used for not only a still image but also a moving image by reducing the cost and size of the applied device.

Further, according to the image signal decoding apparatus of the present invention, the above-mentioned effect can be obtained with the conventional image signal encoding apparatus. That is, the effect can be improved only by devising the decoding apparatus for the standard compression method, and the reproduction can be performed as usual, and the degree of distortion removal can be freely set.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating the concept of an image signal decoding apparatus according to the present invention.

FIG. 2 is a diagram for explaining an operation of a general quantizer.

FIG. 3 is a block diagram of a first embodiment of an image signal decoding apparatus according to the present invention.

FIG. 4 is a block diagram showing a second embodiment of the image signal decoding apparatus according to the present invention.

FIG. 5 is a diagram conceptually showing the operation of the second embodiment.

FIG. 6 is a block diagram showing a third embodiment of the image signal decoding apparatus according to the present invention.

7A is a diagram illustrating orthogonal transform coefficients of a block of interest, and FIG. 7B is a diagram illustrating significant coefficients.

FIG. 8 is a diagram showing a pass band for band cutting.

FIG. 9 is an operation transition diagram for explaining the principle of a conventional image signal encoding and decoding method.

FIG. 10A is a diagram for explaining blocking of image data, and FIG. 10B is a diagram showing a result of discrete cosine transform.

[Explanation of symbols]

10: variable length code decoding circuit, 11: inverse quantization circuit,
12, 16: inverse orthogonal transform circuit, 13: distortion removal circuit, 14
... an orthogonal transform circuit, 15 ... a clipping circuit, 17 ... an output device.

Claims (3)

(57) [Claims] 1. An image data is divided into blocks, an orthogonal transformation is performed for each of the divided blocks, and the transformed output is quantized, and then the quantized output is compressed by performing variable length coding. Variable-length code decoding means for decoding image data, inverse-quantization means for obtaining orthogonal transform coefficients by inverse-quantizing the decoded output from the variable-length code decoding means, and orthogonalization from the inverse quantization means. First inverse orthogonal transform means for performing an inverse orthogonal transform on a transform coefficient, distortion removing means for performing a distortion removing process on the converted output from the first inverse orthogonal transform means, and distortion removing processing from the distortion removing means Orthogonal transform means for obtaining an orthogonal transform coefficient by orthogonally transforming the output, and calculating a change amount of the orthogonal transform coefficient before and after the distortion removal processing for each block from both outputs of the inverse quantization means and the orthogonal transform means. Comparing the calculated amount of change in the coefficient with the maximum value of the possible values of the quantization error of the coefficient,
When the amount of change is larger than the maximum value of the possible values of the quantization error of each coefficient, the orthogonal transform means sets the amount of change to be equal to or less than the maximum value of the possible value of the quantization error of each coefficient. Clipping means for correcting and outputting the orthogonal transform coefficient of the above, and second inverse orthogonal transform means for obtaining a decoded reproduced signal by performing an inverse orthogonal transform on the corrected orthogonal transform coefficient from the clipping means. Image signal decoding device. 2. The apparatus according to claim 1, further comprising means for guiding a reproduced signal from said second inverse orthogonal transform means to said distortion removing means, wherein the distortion removing processing and clipping are repeated at least once. The image signal decoding device according to claim 1. 3. The method according to claim 3, wherein:
3. The image signal decoding apparatus according to claim 2, further comprising: means for changing a distortion removal characteristic of said distortion removing means.