JPH06217284A - Image encoding device - Google Patents

Abstract

PURPOSE:To provide the image encoding device which suppresses variation in quantization between blocks corresponding to two frames adjoining to each other with time and also reduces picture quality deterioration due to the quantization error variation. CONSTITUTION:This image encoding device is equipped with a quantization parameter storage means 15 which stores information regarding quantization step width of each block, used for quantization by a quantizing means 4, block by block, and a quantization parameter control means 14 which controls the quantization step width of each block in a current frame according to the information in the quantization parameter storage means 15 regarding the quantization step width of blocks at the same block in the last frame so that the quantization step width of each block of the current frame is within a certain range centering on the quantization step width of the block at the same position of the last frame.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding device.

[0002]

2. Description of the Related Art An image coding apparatus is provided with an orthogonal transform circuit for orthogonal transforming an image signal or a prediction error signal in block units, a quantizer for quantizing information about transform coefficients by the orthogonal transform circuit, and a quantizer. It is known that a variable length encoder for variable length encoding the quantization level is provided.

In such an image coding apparatus, in order to send the information obtained by the variable length coder at a constant transmission rate, a buffer memory is generally provided at the subsequent stage of the variable length coder. However, if a large amount of information is generated from the variable-length encoder, there is a risk of overflow in the buffer memory. Therefore, the quantization step width of the quantizer is adaptively controlled so that the amount of information generated from the variable length encoder is reduced when overflow is likely to occur. That is, when overflow is likely to occur, the quantization step width is controlled so that the transform coefficient by the orthogonal transform circuit is roughly quantized.

In addition, in an encoding device that suppresses the code amount in a predetermined unit such as several times or a fraction of one frame to a constant amount, the output code amount is estimated from the transform coefficient and the estimated output code amount is obtained. Based on this, the quantization step width of the quantizer is adaptively controlled.

[0005]

As described above, there is an image coding apparatus in which the quantization step width used in the quantizer is controlled by the accumulated amount of the buffer, the estimated amount of the output code amount, etc. No one yet controls the quantization step size for each block of a frame based on the quantization step size used for the co-located block of the previous frame. For this reason, when encoding an image signal, the quantization step width for blocks at corresponding positions in two temporally adjacent frames may fluctuate significantly. Then, since the quantization error fluctuates greatly between both blocks, there is a problem that the image quality deterioration such that the screen is rough is emphasized.

According to the present invention, there is provided an image coding apparatus capable of suppressing a variation in quantization error between corresponding blocks of two frames temporally adjacent to each other and reducing image quality deterioration due to the variation in quantization error. The purpose is to provide.

[0007]

An image coding apparatus according to the present invention comprises an orthogonal transforming means for orthogonally transforming an image signal or a prediction error signal in block units, and a quantizing for quantizing information on transform coefficients by the orthogonal transforming means. Means and a variable length coding means for variable length coding the quantization level obtained by the quantizing means,
Quantization parameter storage means for storing, for each block, information relating to the quantization step width for each block used for quantization by the quantization means, and the quantization step width for each block in the current frame, Based on the information about the quantization step width of the block at the same position of the previous frame stored in the parameter storage means, the quantization step width for each block of the current frame is the quantization step width for the block at the same position of the previous frame. It is characterized in that it is provided with a quantization parameter control means for controlling so as to be within a fixed range with respect to.

When the difference between the feature amounts of the current frame to be quantized and the feature amount of the block at the same position in the previous frame is small or large, the difference between the feature amounts of both blocks is small. Alternatively, the quantization step width control may be performed by the quantization parameter control means.

[0009]

Information on the quantization step width for each block used for the quantization by the quantization means is stored for each block in the quantization parameter storage means. And
The quantization step width for each block in the current frame is based on the information about the quantization step width of the block at the same position in the previous frame stored in the quantization parameter storage means. Are controlled so as to be within a certain range around the quantization step width for the block at the same position in the previous frame.

[0010]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows the present invention in which motion-compensated interframe prediction and two-dimensional DCT (discrete cosine transform) are used.
An example in the case of being applied to a hybrid encoding device in which e transform) is combined is shown.

The input image data is divided into blocks each having a size of 8 × 8 pixels, for example. The input image data of the current frame for each block unit is sent to the subtractor 1. The subtracter 1 calculates the difference between the input image data and the locally decoded data (prediction data) of the previous frame, and outputs the calculation result as prediction error data.

The input image data of the current frame is also sent to the mode determination circuit 13. The mode determination circuit 13 determines whether or not to apply motion-compensated interframe prediction based on the input image data and the prediction error data, and determines which of the interframe prediction application mode and the interframe prediction non-application mode. A mode determination signal indicating whether or not is output. For example, when the image change with respect to the previous frame is large, such as when the scene is switched with respect to the previous frame, the inter-frame prediction non-application mode is selected. This mode determination signal is sent to the switches 2 and 12.

The switches 2 and 12 are interlocked, and when a mode determination signal for selecting the inter-frame prediction application mode is sent, the switches 2 and 12 are switched to the input terminal a side. As a result, the prediction error data output from the subtractor 1 is sent to the DCT circuit 3 via the switch 2. Further, the locally decoded data (predicted data) in the frame memory 10 is sent to the adder 9 via the switch 12.

When a mode determination signal for selecting the inter-frame prediction non-applicable mode is sent to the switches 2 and 12, the switches 2 and 12 are switched to the input terminal b side. As a result, the original image data is sent to the DCT circuit 3 via the switch 2. Further, the data of the value “0” is sent to the adder 9 via the switch 12.

The block-unit data sent to the DCT circuit 3 through the switch 2 is orthogonally transformed based on the DCT, and its transform coefficient is sent to the quantizer 4. The quantizer 4 quantizes the transform coefficient using the quantization step width determined by the quantization parameter control circuit 14. That is, the quantization level is obtained by dividing the transform coefficient by the quantization step width. The quantization level obtained by the quantizer 4 is a variable length encoder (VLC: variable).
It is variable length coded by an able length coding) 5, and is output via the buffer memory 6 together with the motion vector detection data from the motion vector detection circuit 11.

The quantization step size used for quantization is
It is stored in the quantization parameter memory 15 for each block. The quantization parameter memory 15 has a capacity capable of holding a quantization step width for a block for one frame.

Data relating to the degree of accumulation in the buffer memory 6 is input to the quantization parameter control circuit 14,
Basically, the quantization step width is determined based on the storage degree of the buffer memory 6. However, when the quantization step width is determined based only on the degree of accumulation in the buffer memory 6, the variation in quantization error between corresponding blocks of two temporally adjacent frames may increase.

Therefore, the quantization parameter control circuit 14
Limits the range of the quantization step width as follows when determining the quantization step width based on the degree of accumulation in the buffer memory 6. That is, first, of the quantization step widths stored in the quantization parameter memory 15, for the block of the current frame to be quantized, the quantization step width for the block at the same position in the previous frame is set. read out. Then, the quantization step width is controlled so that the quantization step width for the block of the current frame that is the target of the quantization is within a predetermined range around the quantization step width read from the quantization parameter memory 15. .

The data quantized by the quantizer 4 is also sent to the inverse quantizer 7, locally decoded by the inverse quantizer 7, the inverse DCT circuit 8 and the adder 9, and the frame memory 10 is used as prediction data. Memorized in. Frame memory 10
Is input with the motion vector detection data from the motion vector detection circuit 11, and the position of the locally decoded data (predicted data) to be read from the frame memory 10 is corrected according to the detected motion vector. The locally decoded data read from the frame memory 10 is sent to the subtractor 1 as prediction data, and is also sent to the adder 9 via the switch 12 for local decoding.

By the way, when the scene is different from the previous frame, it is not necessary to adjust the quantization step width between the corresponding blocks of the current frame and the previous frame. Therefore, in such a case, the quantization step width for the block of the current frame may be determined based on only the accumulation degree of the buffer memory 6 without limiting the range of the quantization step width.

That is, based on the transform code for each block output from the DCT circuit 3, a block feature amount extraction means for extracting the feature amount of the block and a feature amount storage for storing the extracted feature amount for each block. And means are provided. Then, when quantizing the transform coefficient of each block of the current frame, the feature amount of the block of the current frame and the feature amount of the block at the same position of the previous frame are read out from the feature amount storage means and compared, and both feature amounts are read. When the difference between is small,
The quantization step width for the block of the current frame is limited within a predetermined range with the quantization step width for the block at the same position in the previous frame as the center. When there is a large difference between the feature amount of the block of the current frame and the feature amount of the block at the same position in the previous frame, the range of the quantization step width is not limited and only the accumulation degree of the buffer memory 6 is used for the current frame. Determine the quantization step size for the block of.

In this case, the conversion code itself for each block output from the DCT circuit 3 is stored in the feature quantity storage means, and based on the content of the storage means, the feature quantity of the block of the current frame and the same position of the previous frame are stored. The feature amount of the block may be obtained and compared. As the feature amount of the block, there are the magnitude of the direct current component representing light and dark, the root mean square of the conversion coefficient representing whether the image is monotonic or complex, and the like.

FIG. 2 shows an embodiment in which the present invention is applied to an information amount prefetch control DCT used for encoding for a digital VTR or the like.

In this information amount prefetch control DCT, image data is encoded in a relatively small unit, for example, a multiple of one frame or a fraction thereof, so that the output code amount falls within a fixed amount. Hereinafter, the unit in which the output code amount should be kept constant is referred to as a fixed length unit.

The input image data is divided into blocks in the formatting memory 21 so as to form fixed length units. The block-converted image data is orthogonally transformed in block units by the DCT circuit 22, and the transform coefficient is output from the DCT circuit 22. The transform coefficient output from the DCT circuit 22 is sent to the quantization parameter control circuit 26 and the adaptive quantizer 24 via the buffer memory 23. The buffer memory 23 is provided for holding the image data until the quantization step width is determined by the quantization parameter control circuit 26.

The adaptive quantizer 24 quantizes the transform coefficient using the quantization step width determined by the quantization parameter control circuit 26. That is, the quantization level is obtained by dividing the transform coefficient by the quantization step width. The quantization level obtained by the quantizer 4 is variable length coded by the variable length encoder (VLC) 25 and output.

The quantization step size used for quantization is
Each block is stored in the quantization parameter memory 27. The quantization parameter memory 27 has a capacity capable of holding a quantization step width for a block for one frame.

In the quantization parameter control circuit 26, the DC
The code amount is calculated based on the block-unit conversion coefficient from the T circuit 22, and basically, the quantization step width for each block is determined so that the code amount falls within a fixed amount in a fixed length unit. At this time, in order to suppress a quantization error variation between corresponding blocks of two frames temporally adjacent to each other,
The quantization parameter control circuit 26 limits the range of the quantization step width as follows.

That is, first, of the quantization step widths stored in the quantization parameter memory 27, the current frame block to be quantized is quantized for the block at the same position in the previous frame. Read the step width. Then, the quantization step width is controlled so that the quantization step width for the block of the current frame which is the object of quantization falls within a predetermined range around the quantization step width read from the quantization parameter memory 27. .

Also in this embodiment, the block feature quantity extraction means for extracting the feature quantity of the block based on the transform code for each block output from the DCT circuit 22 and the extracted feature quantity for each block are stored. And a feature quantity storage unit for reading the feature quantity of the block of the current frame and the feature quantity of the block at the same position of the previous frame from the feature quantity storage unit when quantizing the transform coefficient of each block of the current frame. Only when the difference between the two feature quantities is small, the quantization step width for the block of the current frame is within a predetermined range centered on the quantization step width for the block at the same position in the previous frame. The width may be limited.

In this case, the conversion code itself for each block output from the DCT circuit 22 is stored in the feature quantity storage means, and based on the content of the storage means, the feature quantity of the block of the current frame and the same position of the previous frame are stored. The feature amount of the block may be obtained and compared.

FIG. 3 shows an embodiment in which the present invention is applied to a buffer control DCT encoder.

The input image data is divided into a plurality of blocks in the frame memory 31, and is output from the frame memory 31 in block units. The block-unit image data output from the frame memory 31 is the DCT circuit 3
The orthogonal transform is performed in 2, and the transform coefficient is output from the DCT circuit 32. The transform coefficient output from the DCT circuit 32 is sent to the weighting circuit 33, and weighted so that high-frequency components that are not so important to human visual characteristics are reduced.

The weighted transform coefficient output from the weighting circuit 33 is sent to the quantizer 34.
The quantizer 34 quantizes the transform coefficient using the quantization step width determined by the quantization parameter control circuit 37. That is, the quantization level is obtained by dividing the transform coefficient value by the quantization step width. The quantization level obtained by the quantizer 4 is the variable length encoder (VL
C) 35 is variable-length coded and output via the buffer memory 36.

The quantization step size used for quantization is
Each block is stored in the quantization parameter memory 38. The quantization parameter memory 38 has a capacity capable of holding a quantization step width for a block of one frame.

The quantization parameter control circuit 37 receives the data regarding the storage degree of the buffer memory 36, and basically determines the quantization step width based on the data regarding the storage degree of the buffer memory 36. However, when the quantization step width is determined based only on the storage degree of the buffer memory 36, the quantization error variation may increase between corresponding blocks of two temporally adjacent frames.

Therefore, the quantization parameter control circuit 37
Limits the range of the quantization step width as follows when determining the quantization step width based on the storage degree of the buffer memory 36. That is, first, of the quantization step widths stored in the quantization parameter memory 38, the quantization step width for the block at the same position in the previous frame is set to the block of the current frame to be quantized. read out. Then, the quantization step width is controlled so that the quantization step width for the block of the current frame that is the target of the quantization is within a predetermined range around the quantization step width read from the quantization parameter memory 38. .

Also in this embodiment, the weighting circuit 33 is used.
The block feature quantity extraction means for extracting the feature quantity of the block, and the feature quantity storage means for storing the extracted feature quantity for each block based on the conversion code for each block output from When quantizing the transform coefficient of each block, the feature amount of the block of the current frame and the feature amount of the block at the same position of the previous frame are read from the feature amount storage means and compared, and when the difference between the two feature amounts is small. Alternatively, the quantization step width may be limited so that the quantization step width for the block of the current frame is within a predetermined range around the quantization step width for the block at the same position in the previous frame.

In this case, the conversion code itself for each block output from the weighting circuit 33 is stored in the feature quantity storage means, and based on the content of the storage means, the feature quantity of the block of the current frame and the same position of the previous frame are stored. The feature amount of the block may be obtained and compared.

[0041]

According to the present invention, the two adjacently in time
It is possible to suppress the fluctuation of the quantization error between the corresponding blocks of one frame to be low, and it is possible to reduce the deterioration of the image quality due to the fluctuation of the quantization error.

[Brief description of drawings]

FIG. 1 is an electrical block diagram showing an embodiment when the present invention is applied to a hybrid encoding device that combines motion-compensated interframe prediction and two-dimensional DCT.

FIG. 2 is an electrical block diagram showing an embodiment in which the present invention is applied to a look-ahead control DCT encoding device.

FIG. 3 is an electrical block diagram showing an embodiment when the present invention is applied to a buffer control DCT encoding device.

[Explanation of symbols]

 3, 22, 32 DCT circuit 4, 24, 34 Quantizer 6, 36 Buffer memory 14, 26, 37 Quantization parameter control circuit 15, 27, 38 Quantization parameter memory 5, 25, 35 Variable length encoder (VCL )

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification code Office reference number FI technical display location H04N 1/417 9070-5C 7/137 Z

Claims (2)

[Claims] 1. An orthogonal transforming means for orthogonally transforming an image signal or a prediction error signal in block units, a quantizing means for quantizing information about transform coefficients by the orthogonal transforming means, and a quantizing level obtained by the quantizing means. In a picture coding apparatus equipped with a variable length coding means for variable length coding, the quantization step width for each block, which is used for the quantization by the quantizing means, is stored for each block. The quantization step width for each block in the parameter storage means and the current frame is calculated based on the quantization step width of the block at the same position in the previous frame stored in the quantization parameter storage means. The quantization step width for a block is the quantization step width for the block at the same position in the previous frame. Image encoding apparatus characterized by comprising a quantization parameter control means for controlling to be within a certain range around the flop width. 2. A discriminating means for discriminating whether the difference between the feature amount of the current frame to be quantized and the feature amount of the block at the same position in the previous frame is small or large is provided, and the difference between the feature amounts of both blocks is provided. The image coding apparatus according to claim 1, wherein the quantization step width control is performed by the quantization parameter control means only when is small.