JP2003018600A - Image decoding apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve the problems where, in a conventional MPEG image decoding apparatus, a low-pass filter may act even on the part which requires no noise removal because filtering is controlled based on a motion information for high quantization precision, and the block distortion is promoted by quantization error to degrade image quality if an edge enhancement filter is used. SOLUTION: While predicting appearance of block distortion and mosquitos with the use of the quantization information at encoding, a low-pass filter or an edge enhancement filter is appropriately selected to act on a decoded image, to raise quality of an output image.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique effectively applied to a decoding device for an image signal compressed by an image encoding system such as MPEG.

[0002]

2. Description of the Related Art In a conventional image decoding device, a low pass filter is generally used to reduce noise contained in an output image. For example, in an image encoding system such as MPEG, the encoding efficiency deteriorates when the motion of the image is large. Therefore, in order to reduce the generated code amount, the quantization is roughened at the time of encoding, and as a result, distortion often occurs in each block which is a unit of encoding (hereinafter referred to as block distortion). In order to reduce the image deterioration due to the block distortion, a method of applying a low pass filter to a decoded image to remove the block distortion is generally well known when the amount of motion of the image is large.

[0003]

In the conventional method of removing block distortion by the low-pass filter, since the low-pass filter is applied for each image frame, the low-pass filter is also applied to a portion where no block distortion occurs. There is a problem that the entire image is blurred.

Since block distortion often occurs when the quantization accuracy at the time of encoding is low, if the quantization accuracy is high, the block distortion is less likely to occur even if the motion amount is large. However, in the block distortion removal method using the low-pass filter that is the conventional method, since only the motion information that is the difference information from the previous image is referred to, the low-pass filter is used even when the quantization accuracy at the time of encoding is high. There is a problem that the whole image is blurred because it is applied.

[0007] Further, an image in which the contour is emphasized is generally recognized as “a better image” than an image in which the contour is dropped, so that the contour needs to be emphasized in the image.
However, the conventional method has a problem that the outline portion of the image is also blurred at the same time because the low-pass filter for removing the block distortion is applied.

An object of the present invention is to reduce block distortion that occurs when a large quantization step is taken in order to reduce the code amount for an image with a large motion, and to prevent the outline of the image from being blurred. Provided is an image decoding device capable of improving display image quality. The above and other objects and novel features of the present invention will be apparent from the description of this specification and the accompanying drawings.

[0007]

The typical ones of the inventions disclosed in the present application will be outlined below. That is, in the present invention, the filtering process using the quantization parameter as the control signal is performed in the subsequent stage of the decoder. In other words, the low-pass filter acts only on the part where block distortion occurs,
The operation of the low-pass filter is controlled by using the quantization information at the time of encoding instead of the motion information. In addition, a filter for emphasizing the contour portion (hereinafter, contour emphasizing filter) is applied to the portion having no block distortion in order to enhance the contour portion of the input image.

As a concrete method for realizing these, a low-pass filter and a contour emphasis filter are provided in the output part of the decoded image, and the low-pass filter and the contour emphasis filter are appropriately switched using the quantization information at the time of encoding. To act on the image.

Further, as a configuration obtained by developing this method,
By using a block distortion removal filter or the like based on quantization information, block distortion is removed before contour enhancement is performed, and then the contour enhancement filter is operated over the entire image to perform contour enhancement with noise suppressed.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows an embodiment of the image decoding apparatus of the present invention. The image filter device 108 of the present embodiment includes a decoded image input terminal 101 and a quantization parameter input terminal 102.
A processed image output terminal 103, a control device 104, a low-pass filter 105, an edge enhancement filter 106, and a selection circuit 107.
The decoded image 117 output by the decoder 109 is used as an input image.

The decoder 109 is a variable length decoder (VLD) 110.
, An inverse quantizer (IQ) 111, and an inverse discrete cosine transformer (IDCT) 112.
13 is decoded by the main decoder 109 and outputs a decoded image 117. At the same time, the table of the quantization parameter used in the dequantization from the dequantizer 111 is given to the control circuit 104 as the quantization parameter signal 118. The image filter device 108 of this embodiment includes a low-pass filter 105, an edge enhancement filter 106, a selection circuit 107 for selecting an image processed by any filter, and the selection circuit 107 based on a quantization parameter value. And a control circuit 104 for controlling.

The decoded image signal 117 input to the decoded image input terminal 101 is a low-pass filter 105 and a contour enhancement filter.
Entered in both 106. At the same time, the control circuit 104 sends the control signal 121 to the control circuit 107 based on the quantization parameter signal 118 of the decoded image, and when the quantization parameter value is large, the low-pass filter output signal 119 is changed to the quantization parameter value. If is smaller, the contour emphasis filter output signal 120 is selected and output from the output terminal 103 as the output image signal 122.

The image decoding apparatus of this embodiment is characterized by the image filter apparatus 108. The image filter device portion will be described below. FIG. 5 shows an example of the low pass filter. This low-pass filter is a 3-tap filter, which is composed of two delay circuits 501 and a product-sum calculator, and input pixels 504 are input to the delay circuits 501 and delayed. Then, the tap coefficient 502 is applied to each output of the delay circuit 501 to add them, and finally the averaging process 503 is performed to obtain an output pixel 505. The filter of the present embodiment passes a low frequency component by multiplying the pixel value at the center by twice the pixel value next to the pixel value and multiplying it by 1/4.

FIG. 6 shows an example of a high-pass filter required in the contour emphasis filter. As in FIG. 5, the filter is a 3-tap filter and is composed of two delay circuits 601 and a product-sum calculator, and the input pixel 604 is input to the delay circuit 601 and delayed. The tap coefficient value 602 of the filter of the present embodiment allows only high-frequency components to pass by taking the difference from the adjacent pixel.

FIG. 7 shows an example of the configuration of the contour emphasis filter 708 using the high pass filter shown in FIG. As shown in FIG. 7, the contour enhancement filter 708 of this embodiment causes the high-pass filter 706 to act on the input image 704, adds the contour of the output to the original input image 704 with an adder, and outputs the output image 7
It can be achieved by getting 05. In order to realize the contour enhancement filter of this embodiment with 3 taps as in the low pass filter, the tap coefficient 702 is set as shown in FIG. This 3-tap contour enhancement filter circuit is composed of a delay circuit 701 and a tap coefficient 702, and is composed of a high-pass filter that allows the components in the high frequency region of the image to pass through and the sum of the original pixel values. The output pixel 705 is obtained by adding a high-pass component to 704 to realize contour enhancement.

FIG. 2 shows one embodiment of the image filter device in the image decoding device of FIG. 1. This image filter device 208 includes a decoded image input terminal 201, a quantization parameter input terminal 202, It is composed of a processed image output terminal 203, a control device 104, a block distortion removal filter 205, a contour enhancement filter 106, and a selection circuit 107.

The quantization parameter signal 218 input to the quantization parameter input terminal 202 is input to the control circuit 104, and the decoded image signal 217 input from the signal decoded image input terminal 201 is first input to the block distortion removal filter 205. Is entered. The control circuit outputs the selection-time control signal 221 according to the quantization parameter value, which selects either the block distortion removal filter output signal 219 or the decoded image signal 217, and after the contour emphasis filter 106 emphasizes the contour, outputs it. The image signal 222 is output.

In FIG. 1, the low-pass filter 205 and the contour enhancement filter 106 are selected, whereas in this method, the cause of noise is removed in advance by the block distortion removal filter 205 before the contour enhancement is performed, and then the contour is removed. Since emphasis is performed, it is possible to always perform contour emphasis. The block distortion removal filter 205 and the contour enhancement filter 106 are controlled by the quantization parameter, as in the case of FIG.

As described above, in this embodiment, the first filter is a one-dimensional low-pass filter and the second filter is a one-dimensional contour enhancement filter. Is a two-dimensional filter,
Alternatively, another filter such as a noise reducer can be used, and it is easy to improve the image quality by controlling these filters using the quantization parameter value. Can be analogized to.

The image filter device of FIG. 3 is a development of the method of FIG. 1, and this image filter device 308 has a decoded image input terminal 301, a quantization parameter input terminal 302, a processed image output terminal 303, and a control. Device 104 and low pass filter 30
5, contour enhancement filter 306, selection circuit 107, contour enhancement filter table 309, and low-pass filter table 3
It consists of 10.

This filter table has the tap coefficient of the low-pass filter 105 in the low-pass filter table 310 and the tap coefficient of the contour enhancement filter 306 in the contour enhancement filter table 309 in the form shown in FIG. Therefore, the greatest feature is that the tap coefficient value of each filter is selected according to the value of the quantization parameter. Decoded image signal 317 input to input terminal 301
Is input to the low-pass filter 105 and the contour enhancement filter 106.

Further, the quantization parameter signal 318 is input to the control circuit 104 from the quantization parameter input terminal 302. With this quantization parameter value, the contour emphasis filter table control signal 323 and the low pass filter table control signal 324 are output to the contour emphasis filter table 309 and the low pass filter table 310, respectively. Then, by this control signal, the edge enhancement filter tap coefficient
325 and the low pass filter tap coefficient 326 are passed to each filter circuit. In addition, the control circuit 107 selects either the low-pass filter output signal 319 or the contour emphasis filter output signal 320 according to the selection circuit control signal 321 from the control circuit 104, and sets it as the output image signal 322.

FIG. 8 shows a configuration example of a filter for changing the tap coefficient value of the low-pass filter according to the quantization parameter value. The filter of FIG. 8 has a delay circuit 801.
, A tap coefficient multiplier 802, an averaging coefficient multiplier 803,
It is a 5-tap low-pass filter circuit configured with a filter table 806.

In this filter, the input pixel signal 80
Each delay value of 4 has a multiplier 8
After being processed by the tap coefficient in 02, and finally added, all are averaged by the averaging multiplier 803 and become the output image 805. In the filter table 806, two tap coefficient groups are shown as an example. Since the coefficient group on the right side has a narrower band of the filter than the coefficient group on the left side, only the component having a lower frequency is passed. Therefore,
It is possible to analyze the quantization parameter value and select the coefficient on the right side when noise is expected to appear, and select the coefficient on the left side otherwise.

FIG. 9 shows an embodiment of a high-pass filter which is a variable tap coefficient and is used for the contour emphasis filter. The contour emphasis filter circuit of this embodiment is a delay circuit.
This is a 5-tap contour enhancement filter circuit configured by 901, a tap coefficient multiplier 902, an α value multiplier 903, and a filter table 906. The tap coefficient has a narrow band on the right side and a wide band on the left side, as indicated by 906 in the filter table. In this way, by changing the tap coefficient value, the number of taps can be changed.

The filter table 906 is used such that when the contour of the entire image is unsatisfactory, the tap coefficient on the left side with a wide band is used, and on the image with a firm contour, the coefficient group on the right side with a narrow band is used. To be done. In the method shown in FIG. 7, since the contour enhancement is realized by adding the contour portion to the original image, in the embodiment of FIG. 9, by analyzing the quantization parameter, not only the tap coefficient value but also the tap coefficient value is analyzed. , And how much the contour portion, which is the output value of the high-pass filter, is added to the original image. The input pixel value 904 is processed at each tap, and then is multiplied by α in the multiplier 903.
It is multiplied by the value 919 to determine here how many contour components to put on the input pixel value. The α value is controlled using the quantization parameter.

The configuration of FIG. 4 is a development of the method of FIG.
This image filter device 408 has a decoded image input terminal 401,
Quantization parameter input terminal 402 and processed image output terminal 403
A control device 104, a block distortion removal filter 205,
A contour enhancement filter 406, a selection circuit 107, a contour enhancement filter table 409, and a processed image output terminal 403 are provided. The decoded image signal 417 input from the decoded image input terminal 401 is
It is input to the block distortion removal filter 205.

At the same time, a quantization parameter input terminal
The quantization parameter signal 418 is input to 402, and this signal is input to the control circuit 104. By the selection circuit control signal 421 from the control circuit 104, either the block distortion removal filter output signal 419 or the decoded image signal 417 is selected by the selection circuit.
Selected by 107. In this way, the block distortion removal filter 205 is applied to the input image,
By removing the block distortion or the like in advance, the contour emphasis filter 106 in the subsequent stage can be operated over the entire image.

Further, the control circuit 104 includes a filter table 4
The control signal 423 is output to 09 for control, whereby the tap coefficient 425 of the contour enhancement filter 406 is determined and passed to the contour enhancement filter 406. In the contour emphasis filter 406, after applying the tap coefficient 425 to the input image,
The output image signal 422 is output from the processed image output terminal 403.

Although the block distortion removing filter 205 is used in the embodiment of FIG. 4, a median filter is an example of the filter effective for removing block distortion.
As shown in FIG. 10, the median filter is composed of a delay circuit 1001 and a comparator 1006. If the number of taps of the filter is n taps, all n pixel values of the input pixel 1004 are compared, and the pixel value having the median value 1007 is output as the output pixel 1005. The median filter is a filter that can relatively efficiently remove only noise components without losing the original contour of the image.

The most important feature of these filter configurations described above is that they are all controlled by the quantization parameter. One image frame is composed of a set of macroblocks, and the quantization parameter is defined for each macroblock. As shown in FIG. 11, a macroblock is a collection of four DCT blocks 1100 for image compression coding.
The quantization parameter value is constant in units of 1106. Furthermore, this quantization parameter value has a great influence on the image quality.

As shown in FIG. 11, for example, the DCT block includes a DC component 1101, a low frequency component 1102, a vertical high frequency component 1103, a horizontal high frequency component 1104, and a high frequency component 11.
It can be divided into 05. When the value of the quantization parameter of the DC component 1101 is large, the quantization error becomes large in the decoded image, the brightness value of the entire DCT block area changes in DCT block units, and the image is recognized as a block-like image. It becomes block distortion. In addition, the high frequency component 11
If the quantization parameter value in the region of 05 is large,
In addition to the noise called mosquito, which causes distortion in the high frequency range, the lack of high-frequency component information makes it impossible to express sharp contours, which may cause block distortion between adjacent blocks. It is generally known that

The filter device in the present invention is, for example, an image
If you want to filter the whole image,
Filtering is performed in the unit of filter block 1107.
U How to handle quantization parameters in filter block 1107
As an example, as shown in FIG. 11, the filter block 1107
The number of pixels is added to the quantization parameter value of each pixel included in
It can be obtained by overlapping. This value Qav is
Qav = (NL ・ NU ・ Q1 + NR ・ NU ・ Q2 + NL ・ ND ・ Q3 + NR ・ ND ・ Q4) / t²
It is represented by.

According to the method of the present invention, whether or not the occurrence of noise is predicted in the region where the filtering process is performed can be estimated relatively easily and accurately from the Qav value. If the value of Qav is large, the low-pass filter is operated in the method of FIG. 1 and the block distortion removal filter is operated in the method of FIG. Therefore,
In macroblock 1106 with a small quantization parameter value,
A macro block with a large quantization parameter that applies a low-pass filter to enhance the contour without dropping the contour, and 11
In 06, a low-pass filter acts and noise is reduced.

Further, FIG. 13 is considered as an example of a block distortion removing filter for removing only block distortion. The quantization parameter value changes in macroblock units. Therefore, in the filter block 1107 that operates the filter, the existing quantization parameter values are compared in the vertical and horizontal directions, and the maximum value max (Q
1, Q3), max (Q2, Q4), max (Q1, Q2),
If max (Q3, Q4) is taken and max (Q1, Q3) or max (Q2, Q4) is larger than the threshold value, a low-pass filter in the vertical direction is actuated, and max (Q
When 1, Q2) and max (Q3, Q4) are larger than the threshold value, a horizontal low-pass filter is operated.
Furthermore, in this method, a horizontal block boundary detection range 1301 and a vertical block boundary detection range 1302 whose values are variable are set in both the horizontal and vertical directions, and when the macro block boundary 1303 does not exist in this range. Will not detect block boundaries and will not allow the low-pass filter to work.

FIGS. 14 and 15 show the configuration of an embodiment for realizing the method of the present invention shown in FIG. 1 by a circuit. Similarly, FIG. 16 and FIG. 17 show, as a configuration, an implementation example by a circuit of the method shown in FIG. 2 which is an application example of FIG.

FIG. 14 shows an example for realizing FIG. This image filter device 1408 has a decoded image input terminal 14
01, quantization parameter input terminal 1402, processed image output terminal 1403, filter control unit 1404, low-pass filter
1405, contour enhancement filter 1406, selection circuit 107, low-pass filter table 1409, contour enhancement filter table 1410, timing input terminal 1411, low-pass filter control unit 1412, contour enhancement filter control unit 1413, low-pass filter The line memory 1414, the line memory 1415 of the contour emphasis filter, and the selection circuit controller 1416. The line memories 1414 and 1415 are used to store data for several lines of an image when applying a filter. The decoded image signal 1417 is input from the input terminal 1401 and is input to the low pass filter 1405 and the contour enhancement filter 1406.

At the same time, the filter control unit 1404 receives the quantization parameter signals 14 and 14 from the input terminals 1402 and 1411.
18 Timing signal 1419 is input and analyzed. The low-pass filter control unit included in this filter control unit 1404.
1412 and the contour enhancement filter control unit 1413,
The low pass filter control signal 1424 and the contour enhancement filter control signal 1423 are provided to the low pass filter table 1409 and the contour enhancement filter table 1410.

As a result, the tap coefficient 1426 of the appropriate low-pass filter and the tap coefficient 1425 of the contour enhancement filter are
It is passed to each filter circuit. In addition, the selection circuit control circuit 14
The selection circuit control signal 1421 from 16 controls the selection circuit 107 to select either the output of the low pass filter 1405 or the output of the contour enhancement filter 1406, and the output image signal 1422 is output from the processed image output terminal 1403. It In this circuit, the low-pass filter 1405 and the edge enhancement filter
Since the 1406 has its own memories 1414 and 1415, it is possible to operate with relatively little delay.

On the other hand, the image filter device 1508 shown in FIG. 15 has a unified memory using a data bus. This contour enhancement filter circuit is provided with a decoded image input terminal 15
01, quantization parameter input terminal 1502, processed image output terminal 1503, filter control unit 1504, low-pass filter
1505, contour enhancement filter circuit 1506, and selection circuit 107
, Low-pass filter table 1509, contour enhancement / filter table 1510, timing input terminal 1511, low-pass filter control unit 1512, and contour enhancement filter control unit 15
13, a line memory 1514, and a selection circuit controller 1516. The decoded image signal 1517 input from the input terminal 1501 is
It is input to the low pass filter 105 and the contour enhancement filter 106.

Further, the quantization parameter signal 1518 and the timing signal 1519 are inputted to the filter control section 1504 from the input terminals 1502 and 1511, and the low pass filter control section 1512.
Is analyzed by the contour enhancement filter control unit 1513.
The low-pass filter table control signal 1524 and the contour enhancement filter table control signal 1523 are given to the low-pass filter table 1509 and the contour enhancement filter table 1510 from these two control units. As a result, the optimum low-pass filter tap coefficient 1526 and the edge enhancement filter tap coefficient 1526
25 are provided to each filter circuit.

Further, the selection circuit control section 1516 outputs a selection circuit control signal 1512, one of the two signals is selected by the selection circuit 107, and is output as an output image signal 1522 from the processed image output terminal 1203. It In this circuit, the low-pass filter 1505 and the contour enhancement filter 1506 share the memory 1514 via the data bus. For this reason, the circuit scale can be made relatively smaller than in the case of FIG. 12, but the control of the entire circuit becomes slightly complicated.

Next, an example of realizing the circuit of FIG. 2 is shown in FIGS. In FIG. 16, the image filter device 1608
Is a decoded image input terminal 1601, a quantized parameter input terminal 1602, a processed image output terminal 1603, and a filter control unit 16
04, median filter 1605, contour enhancement filter circuit 1606, contour enhancement filter table 409, median filter control unit 1609, and contour enhancement filter control unit
1610, timing input terminal 1611, line memory 1612
And a line memory 1613, and the median filter 1605 is used as the block distortion removal filter 205.

In FIG. 16, the decoded image signal 1617 is input to the input terminal 1601 and its image data is the median
The block distortion and mosquito noise that have been input to the filter 1605 are removed in advance, and then the contour enhancement filter 206
Is operated to output the output image signal 1622. The median filter is the median filter control signal 1624 from the median filter control unit 1609.
Controlled by.

Further, the contour emphasis filter control section 1610 outputs a contour emphasis filter table control signal 1623, which causes the contour emphasis filter tap coefficient 1625 to be outputted from the filter table to the contour emphasis filter 1506. These controls are controlled by the control circuit 1604 which uses the quantization parameter control signal 1618 and the timing signal 1619 as information. In this embodiment, as in the case of FIG. 12, the median filter 1605 is provided with a line memory 1612, and the contour emphasis filter 1506 is provided with a line memory 1613. The feature of the present embodiment is that since each filter is individually provided with a memory, from the input terminal 1601 to the output terminal 1603, filtering processing with a small delay is possible, but on the other hand, since it has an individual memory, The point is that the circuit scale becomes large.

On the other hand, FIG. 17 shows an embodiment of a circuit in which the memories are unified. This image filter device 1708 includes a decoded image control input terminal 1701 and a quantization parameter input terminal 1702.
A processed image output terminal 1703, a filter control unit 1704,
Median filter 1705 and edge enhancement filter circuit 17
06, the contour emphasis filter control unit 1709, the timing signal input terminal 1711, the line memory 1715, and the memory control unit 17
Equipped with 16. The decoded image signal 1717 input to the decoded image input terminal 1701 is first input to the median filter 1705 and subjected to noise removal processing such as block distortion. The timing signal 1719 is input to the timing signal input terminal 1711, and the quantization parameter signal 1718 is input to the quantization parameter input terminal 1702. The contour emphasis filter control unit controls the contour emphasis filter according to the quantization parameter signal 1718.

In this embodiment, the median filter 1705
Operates by directly referring to the quantization parameter value, not by the control unit, and the contour emphasis filter 1706 has a structure that has a filter table in the contour emphasis filter circuit. Optimal filtering processing is performed by the control signal 1723. The output image signal 1722 is output. Further, in this circuit, since the line memory 1715 is shared, complicated processing is required for memory management such as timing control.
These processes are performed by the memory control unit 1716.

The circuit of this embodiment is the same as the circuit of FIG. 13 in that the memories are unified, but the circuit of FIG. 14 accesses the memory because one of the two filters is selected. While there is always one filter, in the present embodiment, the filters are connected in series and the two filters operate simultaneously, so that it takes twice as long to access the memory. Therefore, from the input terminal 1701 to the output terminal
The delay up to 1703 becomes very large, and the increase in the delay value due to the unification of the memories becomes more remarkable as compared with the case of FIG. 14, but the circuit scale can be reduced.

Although the invention made by the present inventor has been specifically described based on the embodiments, the present invention is not limited to the above embodiments, and various modifications can be made without departing from the scope of the invention. Needless to say.

[0050]

As described above, according to the present invention,
When filtering the decoded image, the filter is controlled based on the quantization information, so the effect of the low-pass filter on the part with high quantization accuracy disappears,
The low-pass filter does not act on a portion where block distortion does not exist, and the low-pass filter can act on a portion where block distortion exists. Further, since it is possible to predict the appearance of block distortion or mosquito noise from the quantized information, it is possible to apply the contour emphasis filter to the portion where no noise appears. From the above, it is possible to enhance the contour while suppressing noise, and to significantly improve the image quality.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration of an image decoding apparatus of the present invention.

FIG. 2 is a block diagram showing an embodiment of a filter device in the image decoding device of FIG.

FIG. 3 is a block diagram showing another embodiment of the filter device in the image decoding device in FIG.

FIG. 4 is a block diagram showing an embodiment of the filter device of FIG.

FIG. 5 is a conceptual diagram showing an example of a low-pass filter.

FIG. 6 is a conceptual diagram showing an example of a high pass filter.

FIG. 7 is a conceptual diagram showing an example of a contour enhancement filter.

FIG. 8 is a conceptual diagram showing an example of a low-pass filter having a filter table.

FIG. 9 is a conceptual diagram showing an example of a high-pass filter having a filter table.

FIG. 10 is a conceptual diagram showing an example of a median filter.

FIG. 11 is a conceptual diagram showing an example of region division of a quantization parameter of a DCT block.

FIG. 12 is an explanatory diagram showing an example of how to calculate an average quantized value Qav of a filter block.

FIG. 13 is an explanatory diagram showing an example of a block distortion detection range.

14 is a block diagram showing an embodiment of a filter device in the image decoding device of FIG.

15 is a block diagram showing another embodiment of the filter device in the image decoding device of FIG.

16 is a block diagram showing another embodiment of the filter device of FIG.

FIG. 17 is a block diagram showing another embodiment of the filter device of FIG.

[Explanation of symbols]

101 Decoded Image Input Terminal 102 Quantization Parameter Input Terminal 103 Processed Image Output Terminals 108, 208, 308, 408 Image Enhancement Filter Device 309 Edge Enhancement Filter Filter Table 310 Low-Pass Filter Filter Table 406 Edge Enhancement Filter Circuit 408 Image Enhancement Filter Device 501 delay circuit 502 tap coefficient 503 averaging processing coefficient 708 contour enhancement filter circuits 1408, 1508, 1608, 1708 image filter device 109 decoder 1301 horizontal block boundary detection range 1302 vertical block boundary detection range 1404, 1504, 1604, 1704 filter control Department

Continued front page    (72) Inventor Hiromi Watanabe             3 shares at 6-16 Shinmachi, Ome City, Tokyo             Hitachi Device Development Center F term (reference) 5C059 KK03 KK04 MA00 MA23 MC11                       PP04 PP22 PP25 TA69 TB08                       TC06 TD12 TD15 UA05 UA12                       UA14                 5J064 AA01 BA09 BB14 BC01 BC08                       BC09 BC11 BC16 BC25 BD03

Claims (4)

[Claims] 1. An image decoding apparatus for decoding an image signal compressed by an image encoding method, comprising: a first filter and a second filter, wherein said first filter and second filter are used based on the quantization information executed at the time of encoding. An image decoding device, characterized in that one filter and a second filter are switched and acted on a decoded image. 2. The first filter is composed of a low-pass filter, the second filter is composed of a filter for emphasizing a contour portion of an image, and MP is used as the quantization information.
A quantization scale that encodes a DCT coefficient in EG standard image encoding is used. When the quantization scale is larger than a set value, the first low-pass filter is operated, and when it is smaller, the first low-pass filter is operated. 2. The image decoding apparatus according to claim 1, wherein the second filter is operated to reduce the block distortion occurring in the decoded image and output the decoded image in which the contour of the image is emphasized. 3. The filter coefficient and the number of taps of each of the first filter and the second filter are variable, and the filter coefficient and the tap coefficient are switched based on quantization information. The image decoding device according to item 1. 4. A low-pass filter or a median filter for the purpose of removing block distortion is used as the first filter, and a filter for emphasizing a contour is used as the second filter. It is determined that the first filter acts or does not act based on the quantization information at the time of encoding, and the second filter performs a filtering process on the output image of the first filter. The image decoding device according to claim 1.