JP3536355B2 - Encoding 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 encoding device which gives an optimum quantization width for each block composed of a data area of a predetermined size.

[0002]

2. Description of the Related Art M is a high-efficiency coding method for compressing and coding an image by motion compensation predictive coding and DCT conversion.
PEG (Moving Picture Image Coding Experts Group)
And MPEG2 (Moving Picture Image Coding Experts G
roup Phase 2) is proposed. MPEG and MPEG2
Then, the motion vector is obtained from the error of the current image data subjected to the inverse quantization and the inverse DCT conversion and the previous image data. Motion-compensated predicted image data is generated based on this motion vector, and difference data is obtained from this predicted image data and the next image data. Then, DCT conversion and quantization are performed on this difference data.

FIG. 7 is a block diagram showing an example of an encoding device. The current image data via the input terminal 51 is supplied to the subtractor 52 and the motion vector detector 62. The detection output of the motion vector detector 62 is supplied to the motion compensation predictor 61 having a frame memory and the variable length encoder 55. The subtracter 52 takes the difference between the current image data and the output data of the motion compensation predictor 61. Subtractor 52
Output data is, for example, a two-dimensional DCT of (8 × 8) pixels
DCT conversion is performed by the converter 53. The DCT-transformed image data is quantized by the quantizer 54 and supplied to the variable length encoder 55 and the inverse quantizer 58. Inverse quantizer 58
The image data supplied to is subjected to a process reverse to the process performed by the quantizer 54. The output of the inverse quantizer 58 is inversely DCTed by the inverse DCT converter 59 and then supplied to the adder 60. The predicted image data of the reference frame is supplied from the motion compensation predictor 61 to the adder 60, and these data are added. The output data of the adder 60 is supplied to the motion compensation predictor 61. In the motion compensation predictor 61, the detection output from the motion vector detector 62 and the adder 6
Motion compensated prediction is performed based on the addition output from 0.

The variable length encoder 55, as described above,
The motion vector from motion vector detector 62 and the quantized output from quantizer 54 are provided. Quantizer 5
4 is variable length coded based on the motion vector. The encoded data of the variable length encoder 55 is supplied to the constant rate controller 56. From the constant rate controller 56, encoded data is supplied to the output terminal 57 at a constant rate. Further, the constant rate controller 56 supplies a quantizer scale to the quantizer 54 and the inverse quantizer 58.

[0005]

In the above-described coding device, processing for detecting a motion vector is performed in block units obtained by dividing the entire screen into rectangular areas of a certain size.
Then, using the motion vector, inter-frame motion-compensated prediction in which the current picture block is predicted from the temporally preceding or following picture block, or inter-field motion-compensated prediction is performed. In these motion compensation prediction methods, a motion vector search area having a predetermined size is provided depending on the temporal distance of the predicted screen,
Within the search area, a motion vector that can most accurately perform motion compensation prediction is detected. In order to perform the motion-compensated prediction more accurately, it is desirable to increase the range of possible values of the detected motion vector in the horizontal direction and the vertical direction.

However, in the actual coding apparatus, the range of values that the motion vector can take, that is, the search area is vertically symmetrical or laterally symmetrical depending on the frame, and is uniquely determined with a constant size. Further, the effective search area of the motion vector for each block is not constant regardless of where the current block is located on the screen. That is, when processing blocks close to the four upper, lower, left, and right outer frames of the processing screen, the search areas of the blocks close to the outer frame are reduced as compared with the given search area. A block whose search area has been reduced in this way may be inferior in motion compensation prediction performance to a block whose search area has not been reduced. As a result, the blocks near the outer frame may be deteriorated in the reproduced subjective image quality due to the encoding, as compared with the other blocks. In addition, for example, in an encoding device that divides an entire screen into a plurality of channels and performs parallel encoding processing, the subjective image quality of the reproduced screen is conspicuously deteriorated near the boundaries of the divided screens. May occur.

Therefore, an object of the present invention is to provide an encoding device capable of maintaining the subjective image quality of a reproduced screen above a predetermined quality regardless of the position of a block in the screen.

[0008]

According to the present invention, a DCT converter 3 for compressing data divided into block units, and
A quantizer 4 for quantizing the data DCT-transformed by the CT converter 3, and an effective vector search area ratio detector 1 for detecting a searchable area of a motion vector for the data.
3 and a quantization width determiner 14 that gives a quantization width to the quantizer 4 based on the detection output of the effective vector search area ratio detector 13, and the searchable area is smaller than a predetermined searchable area. At times, the encoding device is characterized in that the quantization width given from the quantization width determiner 14 to the quantizer 4 is reduced.

[0009]

The effective vector search area ratio detector 13 obtains a motion vector searchable area. The quantization width is obtained by the quantization width determiner 14 using the obtained searchable area as one of the factors, and the quantization width is given to the quantizer 4.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An encoding apparatus according to the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an encoding device that performs motion compensation prediction in block units composed of rectangular areas of a certain size on a screen. Input terminal 1
The current image data passed through is supplied to the subtractor 2 and the motion vector detector 12. The motion vector detector 12 is
This is to detect the amount of movement (vector amount) when a block existing at a specific position in the current picture is based on a block of a reference picture that is temporally preceding and succeeding. That is,
As shown in FIG. 2, the block of the current picture 21 is horizontally and vertically referenced to the block of the reference picture 22 (only the previous picture in time in FIG. 2) at the position corresponding to the block of the current picture 21. The motion vector detector 12 detects a vector amount indicating how many pixels are optimally moved in the direction. In addition, mv of FIG. 2 shows a motion vector.

The vector amount detected by the motion vector detector 12 is supplied to a variable length encoder 5 and a motion compensation predictor 11 having a frame memory. The subtractor 2 is supplied with the prediction image data of the motion compensation predictor 11,
The difference between the current image data and the predicted image data is taken. The output data of the subtractor 2 is DCT-converted by the two-dimensional DCT converter 3 of (8 × 8) pixels, for example. Along with this, it is supplied to the effective vector search area ratio detector 13 described later. The DCT-transformed image data is quantized by the quantizer 4 and supplied to the variable-length encoder 5 and the inverse quantizer 8. The image data supplied to the inverse quantizer 8 is subjected to a process reverse to that performed by the quantizer 4. Inverse quantizer 8
The output of is subjected to inverse DCT by the inverse DCT converter 9 and then supplied to the adder 10. The predictive image data of the reference frame is supplied from the motion compensation predictor 11 to the adder 10, and these data are added. The output data of the adder 10 is supplied to the motion compensation predictor 11. The motion compensation predictor 11 performs motion compensation prediction based on the detection output from the motion vector detector 12 and the addition output from the adder 10.

The variable length encoder 5 is supplied with the motion vector from the motion vector detector 12 and the quantized output from the quantizer 4 as described above. As a result, the quantizer 4 performs variable length coding on the quantized output based on the motion vector. The encoded data of the variable length encoder 5 is supplied to the constant rate controller 6. The output data of the constant rate controller 6 is supplied to the output terminal 7 at a constant rate. Further, the quantization scale is supplied from the constant rate controller 6 to the quantization width determiner 14. The effective vector search area ratio detector 13 detects the size of an effective vector detection search area (effective vector area ratio) when the search area of a given vector is set to 1 in the motion vector processing block. Is. The detection output of the effective vector search area ratio detector 13 is supplied to the quantization width determiner 14. The quantization width determiner 14 determines the quantization width based on the quantization scale supplied from the constant rate controller 6 and the effective vector search area ratio supplied from the effective vector search area ratio detector 13. The quantization width output from the quantization width determiner 14 is supplied to the quantizer 4 and the inverse quantizer 8.

By the way, when detecting a motion vector, the magnitudes of the horizontal and vertical components of the vector, that is, the search area of the vector, is determined by the horizontal distance between the current picture and the reference picture. It is uniquely determined symmetrically in the direction and the vertical direction. But,
When the processing block in the current picture is located near the boundary of the screen, the block is reduced in the horizontal direction or the vertical direction as compared with the given search area for vector detection. The degree of the reduction range is uniquely determined by how many blocks the processing block is located from the horizontal or vertical boundary, that is, the number of offset blocks from the horizontal and vertical boundaries. The effective vector search area ratio detector 13 detects a vector searchable area in a block.

Hereinafter, the effective vector search area rate detector 1
The operation in No. 3 will be described with reference to FIGS. 3, 4 and 5.
3, 4, and 5 are diagrams showing the relationship between the block size and the motion vector search range. In FIG. 3, one block has a size of B × B, for example. The search area for vector detection determined by the temporal distance between the current picture and the reference picture is -AREA H to AREA H in the horizontal direction and -AREA V to AREA V in the vertical direction. In order to search for the optimum motion vector, the area around the (B × B) block is set as a search area for motion vector detection. When this block is in an area other than the vicinity of the screen boundary (for example, the center of the screen), the effective vector search area ratio is set to "1".

FIG. 4 is a diagram showing the relationship between the block size and the motion vector search range when the effective vector search area ratio is less than "1". In FIG. 4, one block has a size of (B × B), and (B × B)
B) It is assumed that the block exists near the boundary of the screen (screen edge). In this case, the effective vector search area ratio is less than 1. That is, when this block exists at the edge of the screen, the motion vector detection search area becomes 32. As a result, the effective vector search area ratio becomes less than "1".

FIG. 5 is a diagram showing the relationship between the block size and the motion vector search range when the entire screen is divided into a plurality of channels. In the vicinity of the boundary between the divided screens as shown in FIG. 5, the effective vector search area ratio is less than “1” as in the case shown in FIG.

As shown in FIG. 4 and FIG. 5, the block in the position where the effective vector search area ratio is less than "1" may deteriorate the subjective image quality as compared with the blocks in the other areas. . Therefore, when determining the quantization width and the dequantization width for the quantizer 4 and the dequantizer 8, by supplying the data of the effective vector search area ratio to the quantization width determiner 14, The quantization width can be changed according to the size of the effective vector search area in the block.

In the quantizer 4 in FIG. 1, the pixel value converted by the DCT converter 3 is divided by the quantization width uniquely determined for each block. That is, the pixel value after DCT conversion is pxl, the pixel value after quantization is pxlq, and the quantization width is Δq.
Then, the quantization operation is as follows. pxlq = pxl / Δq (1)

The quantization width Δq is given for each block. This quantization width Δq is a reference quantization width refq obtained by feedback from the buffer control unit that monitors the actual output buffer in order to keep the output rate of the constant rate controller 6 constant, and other factors. It is calculated by the product of K and the following equation. Δq = K × refq (2)

Further, K is the effective vector search area ratio.
Function fu using (VVAR: Valid Vector Area Ratio)
The product of nc (VVAR) and the other factor K1 is obtained by the following equation. K = K1 × func (VVAR) (3)

The quantizer 4 has an effective vector area ratio (V
By giving func (VVAR) in the equation (3) by VAR), the equation (1) is formed. Note that func (VVAR) is a function that satisfies the following expression (4). For example, when there are two effective vector search area ratios VVAR1 and VVAR2, 0 <VV
If AR1 <VVAR2 ≦ 1, func (VVAR1) ≦ func (VVAR2) (4) That is, when the same reference quantization width refq and K1 are given, the smaller the effective vector search area ratio, the smaller the quantum. The conversion width becomes smaller.

FIG. 6 shows the effective vector search area ratio VVA.
9 is a graph showing the relationship between R and the function func (VVAR), for example, the effective vector search area ratio VVAR and the function func when func (VVAR) = 0.5 × VVAR + 0.5 (5)
It shows the relationship of (VVAR). As mentioned above,
Since 0 <VVAR ≦ 1, the function func (VVAR)
Changes with 0.5 <func (VVAR) ≦ 1 ... (6). That is, the function func (VVAR) changes depending on the value of the effective vector search area ratio VVAR,
Eventually, the quantization width will change.

In the above description, the effective vector search area ratio 13 and the quantization width determiner 14 are used, but instead of these, a ROM or the like in which block numbers are stored may be used. In this case, the block number and the effective vector search area ratio corresponding to the block number are stored in the ROM, and the same effect as that of the above-described embodiment can be obtained.

[0024]

According to the present invention, since the quantization width can be changed according to the size of the effective vector search area ratio, the subjectivity of the reproduced screen in the area close to the vertical width or the left and right edges of the screen can be changed. It is possible to prevent deterioration of image quality. Even when the entire screen is divided into a plurality of small screens and each small screen is processed in parallel, it is possible to prevent the subjective image quality deterioration of the reproduced screen near the boundary of each screen. It will be possible.

[Brief description of drawings]

FIG. 1 is a block diagram of an encoding device according to the present invention.

FIG. 2 is a diagram used for explaining a motion vector amount.

FIG. 3 is a diagram showing a relationship between a block size and a motion vector search range.

FIG. 4 is a diagram showing a relationship between a block size and a motion vector search range when the effective vector search area ratio is less than “1”.

FIG. 5 is a diagram showing a relationship between a block size and a motion vector search range when the entire screen is divided into a plurality of channels.

FIG. 6 Effective vector search area ratio VVAR and function func
It is a graph which shows the relationship with (VVAR).

FIG. 7 is a block diagram showing an example of a conventional encoding device.

[Explanation of symbols]

3 DCT converter 4 Quantizer 13 Effective Vector Search Area Rate Detector 14 Quantization width determiner

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68 H03M ^7/ 30-7/50

Claims (2)

(57) [Claims] 1. A compression means for compressing data divided into block units, a quantization means for quantizing the data compressed by the compression means, and an effective means for detecting a searchable area of a motion vector for the data. The vector search area ratio detection means and the quantization width determination means for giving the quantization width to the quantization means based on the detection output of the effective vector search area ratio detection means. An encoding device characterized in that, when the size is smaller than the possible area, the quantization width given to the quantization means from the quantization width determination means is reduced. 2. The encoding apparatus according to claim 1, wherein the quantization width given to the quantization means is changed according to the detection output of the effective vector search area ratio detection means.