JPH08154250A - Moving image coder - Google Patents

Abstract

PURPOSE: To prevent deterioration in image quality caused by non-correlation of time bases of both frames by applying two-way prediction to a frame and coding the frame with zero motion vector when a motion compensation prediction circuit predicts zero motion vector in a picture element block of a 1-way prediction frame. CONSTITUTION: When a picture element block of a 1-way prediction frame is coded as zero motion vector, information representing zero motion vector and a picture element block is stored in a memory 35. On the other hand, a frame memory 21 is accessed via an address generating circuit 30 and surrounding blocks of a reference block are read and stored in block memories B0-B3, and the 2-way prediction with respect to a reference frame used for 1-way prediction by using the stored blocks is processed by referencing the memory 35 as a zero motion vector. Thus, the time base of the same block of 1-way and 2-way prediction is collated with each other and then deterioration in image quality due to time base non-correlation is prevented.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding apparatus used for an image compression encoder, and more particularly, a moving picture of an adaptive inter-frame predictive coding method using one or both of a unidirectional prediction frame and a bidirectional prediction frame. The present invention relates to an image encoding device.

[0002]

2. Description of the Related Art An adaptive inter-frame predictive coding system using a unidirectional prediction frame and a bidirectional prediction frame is disclosed in, for example, Japanese Patent Application Laid-Open No. 2-345, filed by the present applicant.
No. 678 Gazette and Technical Report of the Television Society (ITE
J Technical Report) Vol. Thirteen
No. 60 pp 13-18 (November 30, 1989)
Day). A block diagram of such a conventional moving picture coding apparatus is shown in FIG.

The input image is input to the motion vector detection circuit 1 and stored in the frame memory 2 in advance, and the motion vector is detected between the read image and the read decoded image (reference frame). This detected motion vector is
It is output to the motion compensation prediction circuit 3. Motion compensation prediction circuit 3
Reads the reference block from the frame memory 2 based on the motion vector, obtains the prediction error of the processing block in the input image, and determines the prediction mode for each block. Here, the block means a two-dimensional area of a plurality of pixels, for example, about 8 × 8 or 16 × 16 pixels.

In the case of an intra-frame coded frame, intra-frame coding is performed with a predicted value being a constant value. In the case of a unidirectional prediction frame, there are 1) motion compensation predictive coding 2) interframe predictive coding in which the motion vector is zero 3) intraframe coding in which the predicted value is a constant value, as the prediction mode. In the case of a bidirectional prediction frame, there are 1) forward motion compensation predictive coding 2) backward motion compensation predictive coding 3) bidirectional motion compensation predictive coding 4) intraframe coding in which the prediction value is a constant value. Of the prediction modes of each prediction frame, the optimum prediction mode is determined for each block, and the prediction error, mode and motion vector are output to the encoding circuit 4. The encoding circuit 4 receives the quantization information from the code amount control circuit 5 and performs encoding according to the information. The code obtained by the encoding is output to the output buffer 6. Output buffer 6
Is output to the code amount control circuit 5 to determine the next quantization. The decoding circuit 7 decodes a prediction error from the quantized data and the quantized information. A decoded image is formed by the sum of the decoded prediction error and the motion-compensated prediction value, and is stored in the frame memory 2 as a reference frame.

[0005]

Conventionally, the prediction modes of the bidirectional prediction frame and the unidirectional prediction frame have been individually determined. Therefore, in the unidirectional prediction frame, the motion vector used may not be zero in the block of the bidirectional prediction frame in which the coding mode in which the motion vector is zero is at the same position as the selected block. The lack of correlation of the blocks on the time axis causes the visual image quality to deteriorate.

When the input image has been converted into a composite signal when the component signal is compressed,
This composite signal is YC separated and converted into a component signal. At the time of this YC separation, the color signal in the composite signal mixes with the luminance signal in the moving image portion. Since the phase of the color signal is inverted for each frame, as shown in FIG. 5, when the unidirectionally predicted frames are odd frames, the bidirectionally predicted frame Q1 between them is the same phase as the frame P2, Frame Q2
Has the same phase as the frame P1. Due to the difference in mixing of the color difference signal into the luminance signal due to the color difference component phase difference, the bidirectional prediction mode is hard to be selected, and the bidirectional predicted frame Q1 is predicted from the frame P2 (arrow 10). Q2 is predicted from the frame P1 (arrow 12). When bidirectional prediction is selected, when viewed on the time axis, P1, (P1 + P2) / 2, (P1 +
P2) / 2, P2, so there is little visual discomfort,
When bidirectional prediction is not selected due to the phase difference in the color difference signals, P1, P2 (Q1), P1 (Q2), and P2 alternate, so the correlation on the time axis deteriorates and the visual deterioration occurs. Was invited. In order to avoid this, when the bidirectional prediction mode is uniformly prioritized and the frequency of occurrence of the bidirectional prediction mode is increased, the bidirectional prediction mode, which is the sum of the reference blocks, shows the average value together with the effect of noise removal. Due to the deterioration in the high frequency range due to the low-pass filter due to the removal, the reproduced image was deteriorated.

Generation of prediction error of bidirectional prediction frame
Since the decoded image was used as the reference frame, the coding error that could not be transmitted and that was present in the decoded image was included in the prediction error of the bidirectional prediction frame. Since the coding is performed on the pure motion compensation prediction error and the prediction error in this reference frame, the transmission of the motion compensation prediction error is
There were cases where it was not done enough.

In the unidirectional prediction frame, the prediction error that was not transmitted in the previous unidirectional prediction frame is not transmitted when the quantization accuracy is the same, and the prediction error is stuck to the screen. There was deterioration of the above image quality.

Therefore, it is a first object of the present invention to provide a moving picture coding apparatus that does not cause deterioration of image quality due to the lack of correlation of blocks on the time axis. Further, according to the present invention, when the input image is once converted into the composite signal, the correlation in the time direction of the prediction mode of the bidirectional prediction frame which is likely to be lost due to the phase difference of the color difference signal is calculated for the low resolution region. A second object is to provide a moving picture coding device that can be improved. A third object of the present invention is to provide a moving picture coding apparatus capable of sufficiently transmitting a motion compensation prediction error. A fourth object of the present invention is to provide a moving picture coding apparatus in which a prediction error that has not been transmitted in a previous one-way predicted frame sticks to a screen and causes no deterioration in image quality. .

[0010]

In order to achieve the first object of the present invention, when a motion vector is coded as zero for a predetermined pixel block of a unidirectional prediction frame, the fact and the predetermined pixel are A unit for storing information indicating a block, and when the motion vector is set to zero in the unidirectional prediction frame for a predetermined pixel block in response to the storing unit, the unidirectional prediction frame and the unidirectional prediction frame are used. When bidirectional prediction is performed on a frame between the reference frame and the reference frame, a unit for performing coding with a motion vector of zero for a pixel block located at the same position as a predetermined pixel block is provided.

That is, according to the present invention, in the moving picture coding apparatus of the adaptive inter-frame predictive coding system using the unidirectionally predicted frame and the bidirectionally predicted frame, the motion vector is determined for a predetermined pixel block of the unidirectionally predicted frame. When encoded as zero, a means for storing the fact and information indicating the predetermined pixel block, and the motion vector being zero in the unidirectional prediction frame for the predetermined pixel block in response to the storing means. When the bidirectional prediction is performed for a frame between the unidirectional prediction frame and the reference frame used for the unidirectional prediction, the motion vector is set to zero for the pixel block at the same position as the predetermined pixel block. And a means for performing encoding as a moving picture encoding apparatus.

In the present invention, in order to achieve the second object, the resolution is low in the means for detecting the resolution of the pixel block to be processed and the means for detecting the resolution when the bidirectional prediction frame is encoded. For a pixel block determined to be high, a means for increasing the priority of bidirectional prediction is provided.

That is, according to the present invention, in the moving picture coding apparatus of the adaptive interframe predictive coding system using the bidirectional prediction frame, the resolution of the pixel block to be processed when the bidirectional prediction frame is coded. And a means for increasing the priority of bidirectional prediction for a pixel block whose resolution is determined to be low by the means for detecting resolution. Provided.

According to the present invention, in order to achieve the third object, the reference pixel block used for encoding a bidirectional prediction frame is composed of decoded image data after encoding and original image data before encoding. A means for adaptively switching and reading from any one is provided.

That is, according to the present invention, in the moving picture coding apparatus of the adaptive interframe predictive coding system using the bidirectional predictive frame, the reference pixel block used in coding the bidirectional predictive frame is coded. There is provided a moving picture coding apparatus having a means for adaptively switching and reading from either decoded image data after coding or original image data before coding.

In the present invention, in order to achieve the fourth object, a reference pixel block used in encoding a unidirectional prediction frame is composed of decoded image data after encoding and original image data before encoding. Means for reading from both, and means for changing the quantized coefficient used for encoding depending on the difference between the decoded image data and the motion compensation prediction error using the reference block by the original image data are provided.

That is, according to the present invention, in the moving picture coding apparatus of the adaptive interframe predictive coding system using the unidirectional prediction frame, the reference pixel block used in the coding of the unidirectional prediction frame is coded. A means for reading from both the decoded image data after encoding and the original image data before encoding and the difference between the decoded image data and the motion compensation prediction error using the reference block by the original image data are used for encoding And a means for changing a quantization coefficient.

[0018]

1 is a block diagram of a preferred embodiment of a moving picture coding apparatus according to the present invention. The input image is input to the motion vector detection circuit 20. At this time, in the case of the intra-frame coded frame and the unidirectional prediction frame, the input image is stored in the frame memory 21. In the processing of the intra-frame coded frame, coding is performed with the predicted value as a constant value as in the conventional case, and the coded image is output to the frame memory 21. In the case of a unidirectional prediction frame, a motion vector (MV) is detected by the motion vector detection circuit 20 between the original image of the reference frame accumulated in the frame memory 21 and the input image. The detected motion vector is output to the motion compensation prediction circuit 22. The motion compensation prediction circuit 22 receives the quantization information from the code amount control circuit 23, and determines the quantization coefficient and the prediction mode. The determined prediction mode, quantized coefficient and motion vector are
The difference between the prediction value (prediction block) and the processing block is output to the encoding circuit 24 and is output to the encoding circuit 24. The encoding circuit 24 encodes the prediction error using the quantized coefficient, outputs the code to the output buffer 25, and in the case of the intra-frame coded frame and the unidirectional predicted frame, the quantized prediction error and the quantized error. The coefficient is output to the decoding circuit 26. The decoding circuit 26 decodes the prediction error from the quantized prediction error and the quantized coefficient, and outputs the sum of the prediction error and the predicted value to the frame memory 21 as a decoded image. The output buffer 25 outputs the state of the buffer to the code amount control circuit 23, and the code amount control circuit 23 generates information used for the next quantization.

FIG. 2 is a detailed block diagram of the motion compensation prediction circuit in FIG. In the motion compensation prediction circuit 22,
Based on the detected motion vector (zero vector and input vector), the address generation circuit 30 uses the frame memory 2
The address of 1 is generated and output, the data around the reference block is read, the block is generated using the pixel interpolation circuit 31, and the blocks are sequentially stored in the block memories (B0, B1, B2, B3). The block read at this time inputs the data of the data B2 read from the original image based on the motion vector, the data of the data B0 read from the decoded image based on the motion vector, and the zero vector to the address generation circuit 30, There are four types of data, B1 data read from the decoded image with the vector being zero, and B3 data read from the original image with the motion vector being zero. On the other hand, the processing block input via the motion vector detection circuit has its DC value calculated by the DC calculation circuit 32, and is supplied to the mode determination circuit 33.

FIG. 3 is a detailed block diagram of the mode decision circuit in FIG. The input processing block is stored in the processing block memory 41. The sequencer 42 converts the block input to the multiplication circuit 43 into the DC component input from the DC operation circuit 32, the block memory B0, the block memory B1, the block memory B2, and the block memory B3.
A switch control signal for switching in this order is created and given to the switch 47. The sequencer 42 also sends out the address signals of the block memories B0 to B3. Switch 47
Each pixel of the processing block is read from the processing block memory 41 with respect to each of the signals sequentially selected by the switching by, and the square error of each pixel is calculated, and the result is output to the determination circuit 44. In the case of the square error with the DC component, the difference between the input block and the constant value is stored in the prediction block memory 45. The square error of the data of the block memory B0 and the square error of the DC component are compared by the determination circuit 44, and when the data of the block memory B0 is selected, the data of the block memory B0 is written to the prediction block memory 45. . The comparison with the data in the block memory B1 is performed in the same manner. At this stage, when the prediction block based on the motion vector of the data in the block memory B0 is selected, the data in the block memory B2 is read out, the squared error is calculated with the processing block, and the result is output to the determination circuit 44. The determination circuit 44 compares the square error between the data in the block memory B0 and the data in the block memory B2. If the square error value of the data in the block memory B2 is small and the square error value of the data in the block memory B0 is large, Since it is considered that a large amount of quantization error other than motion compensation prediction error is mixed in the reference decoded image,
The quantization level input from the code amount control circuit 23 is changed by the quantization coefficient circuit 46 so that the quantization is finely performed. Further, when the data in the block memory B1 is selected, the reference block from the original image is read from the block memory B3 in the same manner, the necessity of changing the quantization coefficient is obtained, and the NOMC (NO MOTION-COMPENSATION) memory 35 is also stored. Write 1 or 0 if not selected.

The prediction block is a decoded value of the prediction error output from the encoding circuit 24, that is, the decoding circuit 2.
When the output data of 6 is output, the motion compensation prediction circuit 22
The decoded image is output, and the sum of the two, which is a decoded image, is stored in the frame memory 21.

In the case of a bidirectional predictive frame, there are two directions (forward) and backward (future) between the original image of the reference frame stored in the frame memory 21 and the input image in the motion vector detection circuit 20. The motion vector detection is performed, and the detected motion vector is output to the motion compensation prediction circuit 22. The motion compensation prediction circuit 22 generates an address of the frame memory by the address generation circuit based on the forward motion vector, reads the data from the frame memory 21, and blocks the pixel interpolation circuit 31 based on the motion vector. Is generated and written in the block memory B0. Similarly, based on the backward motion vector, the frame memory 21
More data is read, a block is generated by the pixel interpolation circuit 31, written in the block memory B1, a block in the forward direction is read from the data in the block memory B0, and a bidirectional prediction block is generated by the adder 36. Write to block memory B2. The input block to the address generation circuit 30 is switched to zero, the prediction block when the motion vector is set to zero is similarly read, the forward block is stored in the block memory B3, and the backward block is read from the frame memory 21. The average value with the data in the read block memory B3 is obtained by the adder 36 and written in the block memory B3.

At this time, if there is a margin in the transmission line due to the state of the transmission line input by the code amount control circuit 23,
It is possible to select the decoded image in the frame memory as the reference image to be read and switch to the original image when there is no room. As a result, when the coding error included in the decoded image does not have enough room in the transmission path, the coding error is tried to be forcibly transmitted and the quantization is roughened, which can improve the prediction error. The switch 37 switches the output data of the pixel interpolating circuit 31 and the adder 36 in response to the switch control signal from the address generating circuit 30 and supplies it to the block memories B0 to B3. The control of the switch 37 is performed according to a program stored in a ROM (not shown) in the address generation circuit 30 so that the switch can be switched for each predetermined frame.

In the processing block memory 41, the DC value of the processing block is calculated in the DC calculation circuit 32 in the same manner as in the one-way prediction frame, and is output to the mode determination circuit 33 together with the DC component and the resolution calculation is performed. The circuit 34 calculates the resolution of the processing block. The calculation is performed using, for example, the sum of absolute values between adjacent pixels in the processing block. The sum of absolute values obtained is the mode determination circuit 33.
Is output to

In the decision circuit 44, as in the case of the unidirectional prediction frame, the processing block and its DC value are used, and the multiplication circuit 43 calculates the sum of squares of the values obtained by subtracting the DC component from the pixels in the processing block. A block obtained by subtracting a constant value from the processing block is written in the prediction block memory 45. Hereinafter, reference blocks are read out from the block memories B0, B1, B2, and B3, and are sequentially processed blocks.
The calculation of the squared error is performed in the multiplication circuit 43, and the calculation result is output to the determination circuit 44. Further, when the unidirectional prediction frame is encoded, from the written NOMC memory 35,
The value of the corresponding block is read and input to the determination circuit 44.

The determination circuit 44 determines the mode by using the sequentially input multiplication results. When the input of the resolution calculation circuit 34 is less than a certain value, the resolution of the processing block is
If the value is low, the multiplication result of the data in the bidirectional prediction block memories B2 and B3 is multiplied by α (0 <α <1) to facilitate bidirectional prediction. Further, when the unidirectional prediction frame is encoded by the data from the NOMC memory 35 with the motion vector set to zero, the bidirectional prediction in which the motion vector of the data in the block memory B3 is set to zero is forcibly selected. Alternatively, or by the same processing as in the case of the resolution calculation circuit 34, bidirectional prediction can be easily performed. The selected mode, the corresponding vector, and the prediction block thereof are output to the encoding circuit 24.

[0027]

As described above, according to the moving picture coding apparatus of the present invention described in claim 1, when the prediction of the unidirectional prediction frame uses the prediction with the motion vector being zero, both Also in the prediction of the directional prediction frame, the prediction with the motion vector set to zero is performed, so that the correlation in the time direction is improved, and thus a visually good decoded image can be obtained. Further, according to the moving picture coding apparatus of the present invention as set forth in claim 2, when the bidirectional prediction frame is coded, the resolution is detected by the means for detecting the resolution of the pixel block to be processed and the means for detecting the resolution. For a pixel block that is determined to be low, a means for increasing the priority of bidirectional prediction is provided. Therefore, if the input image is once converted into a composite signal, it is likely to be lost due to the phase difference of the color difference signal. The correlation of the prediction modes of the prediction frame in the time direction can be improved in the low resolution region. Further, according to the moving picture coding apparatus of the present invention as set forth in claim 3, the reference pixel block used at the time of coding the bidirectional prediction frame is the decoded image data after coding and the original image data before coding. Since a means for adaptively switching and reading from any of the above is provided,
The motion compensation prediction error can now be transmitted sufficiently. Further, according to the moving picture coding apparatus of the present invention as defined in claim 4, the reference pixel block used when coding the unidirectional prediction frame is the decoded image data after coding and the original image data before coding. And a means for changing the quantized coefficient used for encoding according to the difference between the decoded image data and the motion-compensated prediction error using the reference block based on the original image data. The prediction error that was not transmitted in the directional prediction frame no longer sticks to the screen and causes deterioration of image quality. .

[Brief description of drawings]

FIG. 1 is a block diagram of a preferred embodiment of a moving picture coding apparatus according to the present invention.

2 is a block diagram showing a preferred example of a motion compensation prediction circuit in the moving picture coding apparatus of FIG. 1. FIG.

FIG. 3 is a block diagram showing a preferred example of a mode determination circuit in the motion compensation prediction circuit of FIG.

FIG. 4 is a block diagram of a conventional moving image encoding device.

FIG. 5 is a schematic diagram for explaining an operation in the conventional moving image encoding device in FIG.

[Explanation of symbols]

20 motion vector detection circuit 21 frame memory 22 motion compensation prediction circuit 23 code amount control circuit 24 coding circuit (address generation circuit 30, pixel interpolation circuit 31, mode determination circuit 33, adder 36, switch 37, block memory B0
~ B3 constitutes a means for encoding with a motion vector as zero) 25 output buffer 26 decoding circuit 30 address generation circuit 31 pixel interpolation circuit 32 DC operation circuit 33 mode determination circuit (higher priority of bidirectional prediction) Means, means for adaptively switching and reading, means for reading from both, means for changing the quantization coefficient) 34 resolution calculation circuit (means for detecting resolution) 35 NOMC memory (that the motion vector is coded as zero, Means for storing information indicating the pixel block) 36 adder 41 processing block memory 42 sequencer 43 multiplication circuit 44 determination circuit 45 prediction block memory 46 quantized coefficient circuit B0, B1, B2, B3 block memory

Claims (4)

[Claims] 1. A moving picture coding apparatus of an adaptive inter-frame predictive coding system using a unidirectional prediction frame and a bidirectional prediction frame, wherein a motion vector is coded as zero for a predetermined pixel block of the unidirectional prediction frame. When that fact and the means for storing information indicating the predetermined pixel block, in response to the means for storing, when the motion vector is zero in the unidirectional prediction frame for the predetermined pixel block, When bidirectional prediction is performed on a frame between the unidirectional prediction frame and a reference frame used for the unidirectional prediction, a motion vector is set to zero for a pixel block at the same position as the predetermined pixel block and encoding is performed. And a means for performing the moving picture coding apparatus. 2. A moving picture coding apparatus of an adaptive interframe predictive coding system using a bidirectional prediction frame, comprising means for detecting the resolution of a pixel block to be processed when the bidirectional prediction frame is coded. A moving picture coding apparatus comprising: a means for increasing the priority of bidirectional prediction for a pixel block whose resolution is determined to be low by the means for detecting the resolution. 3. A moving picture coding apparatus of an adaptive inter-frame predictive coding system using a bidirectional prediction frame, wherein a reference pixel block used in coding the bidirectional prediction frame is a decoded image after coding. A moving picture coding apparatus having means for adaptively switching and reading out from either data or original picture data before coding. 4. A moving picture coding apparatus of an adaptive interframe predictive coding system using a unidirectional prediction frame, wherein a reference pixel block used in coding the unidirectional prediction frame is a decoded image after coding. Means for reading from both the data and the original image data before encoding, and the quantization coefficient used for encoding is changed by the difference between the decoded image data and the motion compensation prediction error using the reference block by the original image data. A moving image coding apparatus, comprising: