JPH08181990A - Image decoder - Google Patents

Abstract

PURPOSE: To eliminate mosquito noise by providing a noise elimination control means providing a characteristic of noise elimination adaptively for each block based on block type information decoded by a variable length decoding section so as to prevent a fogged image through weakening filter capability when a quantization error is small and through a strengthened filter capability when large. CONSTITUTION: A noise elimination control circuit 8 receives block type information from a variable length decoding circuit 1 to detect a quantization step width to control adaptively a noise elimination processing circuit 6. That is, when the quantization error is small, that is, a quantization step width is small, a threshold level of a sigma filter used for a filter to eliminate mosquito noise is selected small so as to weaken the filter function thereby preventing a fogged image. As a result, the preservation of an edge is attained. When the quantization error is large, that is, a quantization step width is large, the threshold level is selected large to strengthen the filter function thereby eliminating mosquito noise.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image decoding apparatus for quantizing a block-divided pixel image after orthogonal transformation and further decoding a variable length coded image signal into a pixel image.
In particular, the present invention relates to an image decoding device that removes noise from a pixel image after decoding.

[0002]

2. Description of the Related Art Generally, an image decoding apparatus has a large amount of image information, and particularly a moving image has a huge amount of data. Therefore, in the conventional image decoding apparatus, it is necessary to perform some kind of highly efficient compression in order to transmit or record such image information. As a highly efficient compression method of image data, a method of achieving high compression using orthogonal transform coding is generally used. An example of this method will be described below with reference to FIG.

First, the input image signal A is blocked by the blocking circuit 11
Is divided into predetermined blocks, and the orthogonal transformation circuit 12 performs orthogonal transformation on the divided blocks to generate a frequency component signal B. As this orthogonal transformation, for example, 2
Dimensional Fourier transform, Walsh-Hadamard transform, Karhunen-L
There are oeve transform, discrete cosine transform (DCT), and the like. Next, the orthogonal transformation data is linearly quantized by the quantization circuit 14 with the quantization step width (QS) set by the quantization step width control circuit 13. Further, the quantization result Bq is assigned a variable length code by the variable length coding circuit 15 to generate a coded image signal C.

FIG. 5 shows a block diagram of a general image decoding apparatus. This decoding device generates a reproduced image signal A ′ corresponding to the input image signal A from the encoded image signal C generated by the image encoding device of FIG. The transmitted encoded image signal is subjected to variable length decoding by the variable length decoding circuit 21, and then the inverse quantization circuit 22 performs inverse quantization with the quantization step width (QS) obtained by the variable length decoding unit 21. To convert. Further, the inverse orthogonal transform 23 obtains the reproduced image signal A ′. At this time, since the orthogonal transformation and the quantization are irreversible transformations, the image signal A ′ obtained by the image decoding device includes an error, and particularly, the quantization error due to the quantization / inverse quantization deteriorates the image quality of the reproduced image. Is causing
The larger the quantization step width, the larger the quantization error and the more noticeable the deterioration of the image quality of the reproduced image signal. The quantization error of the transform coefficient appears as block distortion in which a discontinuity occurs at a block boundary portion in a reproduced image, or mosquito noise in which a flat shape is visible in a flat portion near an edge.

Therefore, it has been proposed to apply a low-pass filter (low-pass filter) 24 to the reproduced image reproduced by the decoding device for noise removal processing.

[0006]

However, when the image or the like includes edges in the low pass filter, the image is blurred, and when the degree of low frequency passing through the low pass filter is loosened, noise is completely eliminated. There was a problem that it could not be removed.

The present invention has been made to solve the above-mentioned conventional problems, and is an image decoding apparatus for quantizing a block-divided image signal after orthogonal transformation and further decoding the coded image signal into an image signal. In the above, it is an object of the present invention to provide an image decoding device capable of removing noise in a reproduced image after decoding.

[0008]

The present invention is to solve these problems, and divides image data into blocks, and uses orthogonal transformation, quantization, and variable length coding for each of the divided blocks. In an image decoding apparatus for decoding compressed image data, a variable length decoding means for variable length decoding the compressed image data, and an inverse quantum for inverse quantizing the decoded output from the variable length decoding means. Conversion means, inverse orthogonal transformation means for performing inverse orthogonal transformation on the inverse quantized output from the inverse quantization means, noise removal processing means for removing noise from the output from the inverse orthogonal transformation means, and the variable Provided is an image decoding device having noise removal characteristic control means suitable for noise removal characteristics based on block type information obtained from long decoding means.

[0009]

According to the present invention, the characteristic of the noise removing circuit for performing noise removing processing on the reproduced image obtained by performing inverse quantization and inverse orthogonal transformation on the decoded output from the variable length decoding means is described above. It is changed based on the block type information obtained by the variable length decoding means, particularly the quantization step width.

[0010]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of an image decoding apparatus according to the present invention. In the figure, reference numeral 1 denotes a variable length decoding circuit for decoding coded data for each block of an image. Then, it is output to the addition circuit 4 via the inverse quantization circuit 2 and the inverse orthogonal transformation circuit 3. The output of the adder 4 is added to itself through the motion compensation prediction circuit 5 and is supplied to the image output circuit 7 through the noise removal circuit 6. The noise removal control circuit 8 receives the block type information from the variable length decoding circuit 1 to detect the quantization step width, and adaptively controls the noise removal processing circuit 6. The operation of the configuration of FIG. 1 will be described.

Since the orthogonal transformation and the quantization are irreversible transformations, the image signal obtained by the image decoding apparatus contains an error. Especially, the quantization error due to the quantization / inverse quantization causes the deterioration of the image quality of the reproduced image. Therefore, the larger the quantization step width, the larger the quantization error and the more noticeable the deterioration of the image quality of the reproduced image signal. The quantization error of the transform coefficient appears as block distortion in which a discontinuity occurs at a boundary portion of a block in a reproduced image or mosquito noise in which a haze is seen even in a flat portion near an edge. A sigma filter is used as a filter for removing this mosquito noise. First, the sigma filter will be described. The sigma filter takes the difference between the target pixels for ± 3 pixels in the vicinity of the target pixel, and when the absolute value is smaller than the threshold value, weights the difference and adds it to the target pixel. By this filter, the muffled noise appearing on the flat portion is reduced. Therefore, when the quantization error is small, that is, when the quantization step width is small, the mosquito noise is less likely to occur, so the threshold value is made smaller to weaken the filter strength and prevent blurring of the image. Edge preservation is also possible. When the quantization error is large, that is, when the quantization step width is large, the threshold value is increased to increase the filter strength and remove the mosquito noise. The noise removal control circuit will be described below with reference to FIG.

The block type information input from the variable length decoding circuit 1 of FIG. 1 is detected by the quantization step width detecting circuit 9 and used by the threshold value calculating circuit 10 in the noise removing circuit 6 of FIG. Calculate the threshold.

The characteristics of the noise removal control circuit 8 and the threshold value calculation circuit having the above configuration will be described below with reference to FIG. In the variable length decoding unit of FIG. 1, if the detected quantization step width is within the range of X shown in FIG. 3, the threshold is set to the value of TH1, and if the quantization step width is within the range of Y, the threshold is set. The value is TH2 and the quantization step width is Z
By using a table in which the threshold value is the value TH3 in the range of, the filter characteristic that the S / N ratio becomes highest in the encoded quantization step width can be obtained.

[0014]

As described above, according to the present invention, an image signal obtained by quantizing a block-divided image after orthogonal transformation and further encoding the image signal is applied to each block based on the block type information decoded by the variable length decoding unit. Since it has a noise removal control means that can give a characteristic of noise removal, if the quantization step width is small and the quantization step width is small, the mosquito noise is less likely to occur, so weaken the filter strength. Image blurring can be prevented and edges can be saved. If the quantization error is large, that is, if the quantization step width is large, the filter strength is increased to remove mosquito noise.

[Brief description of drawings]

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

FIG. 2 is a block diagram showing details of a noise removal control unit in FIG.

FIG. 3 is a diagram showing the S / N ratio of the simulation result obtained by the present invention.

FIG. 4 is a block diagram showing a configuration example of a conventional image coding apparatus using a conventional orthogonal transform.

FIG. 5 is a block diagram showing a configuration example of a conventional image decoding device using a conventional orthogonal transform.

[Explanation of symbols]

 1 Variable Length Decoding Circuit 2 Inverse Quantization Circuit 3 Inverse Orthogonal Transformation Circuit 4 Addition Circuit 5 Motion Compensation Prediction Circuit 6 Noise Removal Circuit 7 Image Output Circuit 8 Noise Removal Control Circuit 9 Quantization Step Width Detection Circuit 10 Threshold Detection Circuit 11 Blocking Circuit 12 Orthogonal Transform Circuit 13 Quantization Step Width Control Circuit 14 Quantization Circuit 15 Variable Length Coding Circuit 21 Variable Length Decoding Circuit 22 Inverse Quantization Circuit 23 Inverse Orthogonal Transformation Circuit 24 Low Pass Filter

Claims (1)

[Claims] 1. An image decoding apparatus that divides image data into blocks and decodes the compressed image data using orthogonal transform, quantization and variable length code for each of the divided blocks. Variable-length decoding means for variable-length decoding image data, dequantization means for dequantizing the decoded output from the variable-length decoding means, and inverse orthogonal transformation for dequantization output from the dequantization means An inverse orthogonal transformation means for
From noise removal processing means for removing noise from the output from the inverse orthogonal transformation means, and noise removal characteristic control means suitable for noise removal characteristics based on block type information obtained by the variable length decoding means. An image decoding device characterized by being configured.