JPH0654315A - Method and device for encoding animation picture - Google Patents

Abstract

PURPOSE:To enable high efficient encoding in accordance with scene change by counting to which frame in front and rear respective blocks much refer so as to compensate movement while encoding. CONSTITUTION:A controller reads the frame to be encoded one by one and indicates whether an inter-frame encoding mode or a movement compensating mode. Then, initialization is executed and 600 blocks being the half of a block number in one frame are inputted from a terminal 17. After an encoding processing of all the blocks, the counter 11 of a scene change detecting part 21 inputs the counted value of an inter-block encoding mode to an comparator 12 and compares it with 600 being a threshold value. When the block number of the encoding mode is more than the threshold value, scene change detection is announced from a terminal 18 to the controller, a whole operation is stopped, the frame mode to be encoded is changed from the compensating mode to the inter-frame encoding mode and encoding is re-executed. When the block number is equal to or below the threshold value, the frame contents of a buffer 20 is outputted from the terminal 19.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention finds a motion vector by referring to frames before and after a frame for coding a moving image,
The present invention relates to a moving picture coding method and apparatus for detecting a scene change in moving picture coding for motion compensation.

[0002]

2. Description of the Related Art Conventionally, as a moving image encoding method, encoding using inter-frame correlation has been known. In recent years, a frame is divided into a plurality of blocks and a block similar to the block to be encoded is searched for in the encoded frame, and the displacement of the position is used as a motion vector, and the similar block is encoded as a prediction block. The main method is a motion compensation coding method, in which a difference between and is obtained and is coded together with a motion vector as a prediction error. In particular, since the coding characteristic is good, a method has been considered in which the prediction error is subjected to discrete cosine transform, the obtained coefficient is quantized, and then coded. If a similar block cannot be found or the prediction error is too large,
The image quality is improved by performing coding in the coded block without motion compensation. There is known a method of reducing deterioration due to coding by periodically inserting a refresh frame for performing intra-frame coding with a larger amount of information than that for other motion-compensated frames without motion compensation. . Also, as a means of motion compensation, a frame in which one or more frames are skipped is first encoded, and when the next skipped frame is encoded, already encoded frames are before and after these frames. There is an encoding method in which a motion vector is obtained for both frames to perform motion compensation. These encoding methods are suitable not only for communication but also for storage on media, and are suitable methods for realizing functions such as reverse playback. (Reference: "Technical Description: International Standardization of Video / Audio Coding and Multimedia Synchronization" -Yamada, Onoue (JVC) Vol. 19, No. 4 (199)
0) pp. 236 to 242) FIG. 5 shows a block diagram of an example of hardware implementation. In the figure, 501 and 5
Reference numerals 02 and 513 are frame memories for storing one frame of a moving image. Reference numeral 503 denotes the frame memory 502 for the original blocks sequentially cut out from the frame memory 1.
Is a motion vector detector for searching for a motion vector. A motion compensator 504 generates a prediction block from the frame memory 502 based on the motion vector obtained by the motion vector detector 503 and obtains a prediction error block from the original block. A block evaluator 505 calculates a variance value inside the original block. A block evaluator 506 calculates a variance value inside the prediction error block. Reference numeral 507 is a coding mode determiner that determines the coding mode of the original block from the outputs from the block evaluators 505 and 506. Reference numerals 508 and 509 denote delay devices for temporarily storing blocks and ensuring synchronization. Reference numeral 510 is a delay unit 5 according to the output from the coding mode determiner 507.
This is a selector for selecting the output from 08 and 509. 5
Reference numeral 11 is a coding for inputting and coding in block units the coding mode which is the output of the coding mode determiner 507, the block information which is the output of the selector 510, and the motion vector which is the output of the motion vector detector 503. It is a vessel. 5
Reference numeral 14 is a decoder for once decoding the encoded code. In the moving picture coding for coding the prediction error by performing the motion compensation as described above, the coding efficiency is extremely lowered in the vicinity of a scene change in which the contents of the image are greatly different,
There is a drawback that the deterioration of image quality is noticeable.

[0003]

However, a large-scale hardware is required to calculate the difference for all pixels, and the calculation time is long. Further, even when encoding is performed, since the difference value of all pixels is calculated for all frames, there is a waste of calculation even for frames with no scene change.

Therefore, an object of the present invention is to provide a moving picture coding method and apparatus capable of performing efficient coding in accordance with a detected scene change.

[0005]

In order to solve the above problems, the moving picture coding method and apparatus of the present invention is such that, while performing coding, each block refers to many frames before and after which motion compensation is performed. Is provided, or means for counting the presence or absence of reference to another frame is provided, and means for detecting a scene change from the result is provided. Here, the scene change is determined by comparing the count value with the threshold value, and the threshold value is determined in advance,
Update as encoding proceeds.

[0006]

(First Embodiment) FIG. 1 is a block diagram of a moving picture encoder to which the present invention is applied. In the figure, 1 and 2 are frame memories for storing one frame of a moving image. 3 is a block to be encoded (hereinafter referred to as an original block), which is sequentially cut out from the frame memory 1,
This is a motion vector detector that searches the frame memory 2 to obtain a motion vector. 4 is a frame memory 2 according to the motion vector obtained by the motion vector detector 3.
It is a motion compensator that generates a prediction block from the prediction block and obtains a prediction error block from the original block. Reference numeral 5 is a block evaluator that calculates a variance value inside the original block. A block evaluator 6 calculates a variance value inside the prediction error block. Reference numeral 7 is a coding mode determiner that determines the coding mode of the original block from the outputs from the block evaluators 5 and 6. Reference numerals 8 and 9 are delay devices for temporarily storing blocks to ensure synchronization. Reference numeral 10 is a selector for selecting the outputs from the delay units 8 and 9 according to the output from the coding mode determiner 7. A counter 11 counts the number of coding modes output from the coding mode determiner 7. Reference numeral 12 is a comparator for comparing the count value output from the counter 11 with the threshold value input from the terminal 17 at the time when the encoding of the frame is completed. Reference numeral 13 is an encoding for inputting and encoding, in block units, an encoding mode output from the encoding mode determiner 7, block information output from the selector 10, and a motion vector output from the motion vector detector 3. It is a vessel. Reference numeral 14 is a decoder for decoding the encoded code. A frame memory 15 stores a decoded image. Two
Reference numeral 0 is a buffer that outputs encoded data for each frame. Reference numeral 21 represents a scene change detection unit.

To simplify the explanation of the embodiment, as shown in FIG. 3, a refresh frame for intra-frame coding is inserted every 6 frames, and the other frames move from the previously coded frame in time. I will make compensation. Here, the intra-frame code is subjected to discrete cosine transform in block units, and the obtained transform coefficient is quantized to be encoded. For the sake of explanation, the block size is 8 × 8 and the image size is 320 × 240. An image to be input will be described by taking an example in which luminance and chromaticity are separated.

A control device (not shown) sequentially reads the frames to be encoded one by one, instructs whether the frame mode is the intra-frame encoding mode or the motion compensation mode, and controls the operation of the encoder. There is.

Prior to encoding, the control device also initializes each device. 600, which is half the number of blocks in one frame, is input from the terminal 17.

When the frame to be encoded is in the intra-frame encoding mode, the encoding is performed as follows. Terminal 16
The luminance / chromaticity information of the original image is input from and is stored in the frame memory 1. The coding mode determiner 7 fixes the intra-block coding mode for performing intra-block coding until the coding of the first frame is completed. The counter is not activated. The buffer 20 is left empty. According to an instruction from the control device, an original block is cut out from the frame memory 1 in the main scanning direction, inputted to the selector 10 via the delay device 8. Since the coding mode determiner 7 continues to output the intra-block coding mode, the delay unit 8 is always used in this frame.
Output is selected. The encoder 13 encodes the fact that the mode of this frame is the intra-frame encoding mode at the beginning of the frame and the image size, and outputs the encoded image size to the buffer 20. Following this, a discrete cosine transform is applied to each original block, the obtained transform coefficient is quantized and a code is assigned, accumulated in the buffer 20, and simultaneously output to the decoder 14. The buffer 20 waits for an instruction from the control device and outputs the signal from the terminal 19 to the outside after finishing the coding for all the blocks of the frame. On the decoding side, first, the frame mode and the image size are received and decoded. Decoding is further continued, and as soon as all 8 × 8 coefficients of the quantization result of the discrete cosine transform are accumulated, inverse quantization and inverse discrete cosine transform are performed to obtain a decoded pixel value. The playback position of the block is naturally determined from the image size, and after the last block is decoded, it waits until the next frame mode comes. The same thing is performed by the decoder 14, and the decoded image is stored in the frame memory 15. After that, the control device causes the buffer 20 to start the transmission from the terminal 19, and when the transmission is completed, the buffer 20 informs the control device of the completion of the transmission.
After detecting the end of transmission of the control device, the encoding of the next frame is started.

When the frame to be encoded is in the motion compensation mode, the encoding is performed as follows. Before starting the processing of the frame, the control device moves the contents of the frame memory 15 to the frame memory 2 and clears the frame memory 15. If the immediately preceding frame is in the intraframe coding mode, the fixing of the intraframe coding mode of the coding mode determiner 7 is released. Further, the contents of the buffer 20 are cleared and the counter resets the count value to 0. After that, the luminance / chromaticity information of the original image is input from the terminal 16 to the data of the next frame, and the frame memory 1
To store.

According to an instruction from the control device, original blocks are sequentially cut out from the frame memory 1 in the main scanning method. The processing of each block is described below. The cut out original block is input to the motion vector detector 3. The motion vector detector 3 centers in the original block position (this is uniquely determined by counting the number of blocks) in the image (immediately encoded frame) of the frame memory 2 in the main scanning direction and the sub scanning direction. On the other hand, a block with a high degree of similarity to the region of ± 15 pixels is detected by full search,
Find the motion vector. The obtained motion vector is input to the motion compensator 4. The motion compensator 4 cuts out a prediction block from the frame memory 2 according to this motion vector. At the same time, the original block is input from the frame memory 1, a difference is calculated for each pixel from the prediction block, and a prediction error block composed of this difference value is obtained and output. afterwards,
The original block and the prediction error block are delayed by the delay units 8 and 9, respectively.
Stored in. The data of the original block is the block evaluator 5
First, the average M ₀ within the block is obtained, and then the standard deviation value σ ₀ within the block is calculated. Similarly, the data of the prediction error block is input to the block evaluator 6 and the in-block average M _E and the in-block standard deviation value σ _E are calculated. These standard deviation values σ ₀ and σ _E are input to the coding mode determiner 7. The coding mode determiner 7 determines the coding mode of this block from the magnitude relation of the following equations of these standard deviations.

If σ _E > σ ₀ : coding mode is an intra-block coding mode if σ _E ≦ σ ₀ : coding mode is a motion compensation mode

This coding mode is input to the counter 11 of the scene change detection section 21 and counts the number of blocks which were in the intra-block coding mode. At the same time, the coding mode is also supplied to the selector 10, and when the coding mode is intra-block coding, the original block data which is the output of the delay device 8 is predicted, and when it is the motion compensation mode, the prediction is the output of the delay device 9. Each error block is selected and output. The encoder 13 first encodes the image size as to whether the mode of this frame is the motion compensation frame,
Output to 0. After that, the coding mode of the block to be coded is coded for each block. If the coding mode is the intra-block coding mode, the coding mode of the block is coded, the transform coefficient obtained by performing the discrete cosine transform on the original block is quantized, and the code is assigned. If the coding mode is the motion compensation mode, first, the coding mode of the block and the coding of the motion vector are performed, then the obtained prediction error block is subjected to the discrete cosine transform, and the obtained transform coefficient is quantized. To assign a code. These codes are stored in the buffer 20 and, at the same time, output to the decoder 14. The decoder 14 first receives and decodes the frame mode and the image size as in the refresh frame. Furthermore, the coding mode is decoded for each block, and if the coding mode is the intraframe coding mode, as soon as all 8 × 8 coefficients of the quantization result of the discrete cosine transform are accumulated,
Inverse quantization and inverse discrete cosine transform are applied, and the decoding result 8 ×
As soon as all the coefficients of 8 are accumulated, inverse quantization and inverse discrete cosine transform are performed to obtain a decoding error block. This error block is added to the previously decoded pixel block to obtain a decoded pixel value. The reproduction position of the block is naturally determined from the image size, and after the last block is decoded, it waits until the encoded data of the next frame comes. The same thing is performed by the decoder 14, and the decoded image is stored in the frame memory 15.

After performing the coding process on all the blocks, the counter 11 of the scene change detector 21 inputs the count value of the intra-block coding mode to the comparator 12,
The threshold value 600 set in advance is compared. If the number of blocks in the intra-block coding mode is larger than 600, the control device is notified of the detection of the scene change from the terminal 18. Upon detection of a scene change, the control device temporarily stops the entire operation, changes the mode of the frame to be encoded from the motion compensation mode to the intraframe encoding mode, and then re-encodes the frame. . If the count value is less than or equal to the threshold value, the controller outputs the contents of the frame in the buffer 20 to the outside from the terminal 19.

(Second Embodiment) FIG. 2 is a block diagram of a moving picture encoder to which the present invention is applied. In the figure, 10
Reference numerals 0, 101, 102, and 103 are frame memories for storing one frame of a moving image. A selector 104 selects an output from the frame memories 100 and 101. Reference numeral 105 denotes the frame memory 10 for the original blocks sequentially cut out from the frame memory 100 or 101.
2 is a motion vector detector that searches for 2 to obtain a motion vector. Reference numeral 106 denotes a motion vector detector that searches the frame memory 103 for an original block sequentially cut out from the frame memory 100 or 101 to obtain a motion vector. 107 is the frame memory 10 according to the motion vector obtained by the motion vector detector 105.
This is a motion compensator that generates a prediction block from 2 and obtains a prediction error block from the original block. 108 generates a prediction block from the frame memory 103 according to the motion vector obtained by the motion vector detector 106,
It is a motion compensator that obtains a prediction error block from the original block. Reference numeral 109 is a block evaluator that calculates the sum of squares inside the original block. Reference numerals 110 and 111 denote block evaluators that calculate the sum of squares inside the prediction error block. 112
Is a coding mode determiner that determines the coding mode of the original block from the outputs from the block evaluators 109, 110, and 111. Reference numerals 113, 114, and 115 are delay devices for temporarily storing blocks to ensure synchronization. 116
Is a selector that selects the output from the delay devices 113, 114, and 115 according to the output from the coding mode determiner 112. A selector 117 selects the output from the motion vector detectors 106 and 107 according to the output from the coding mode determiner 112. A counter 120 counts the number of coding modes output from the coding mode determiner 112 for each mode. Reference numeral 121 is a comparator that compares the count value output from the counter 120 with a threshold value at the time when the coding of the frame is completed. Reference numeral 122 is a threshold value determiner that determines a threshold value from the count value of the counter 120. Reference numeral 119 denotes an encoder for inputting and encoding, in block units, the encoding mode output from the encoding mode determiner 112 and the block information and motion vector information output from the selectors 116 and 117. Reference numeral 118 is a decoder for decoding the encoded code. Reference numeral 127 represents a scene change detection unit.

To simplify the description of the embodiment, as shown in FIG. 4, a refresh frame for intra-frame coding is inserted every 30 frames. For the remaining frames, motion compensation is performed only from the odd-numbered frame immediately before the odd-numbered frame. Therefore, odd-numbered frames are called interlaced coded frames. This interlaced encoded frame is encoded before the immediately preceding even-numbered frame. It is possible to refer to both the even-numbered frames because the preceding and following frames have already been encoded. Therefore, even-numbered frames are called bidirectionally predicted frames. Motion compensation is performed from encoded frames that are temporally before and after. Further, here, the intra-frame code is subjected to discrete cosine transform in block units, and the obtained transform coefficient is quantized to be encoded. The block size is 8 × 8 for the sake of explanation, and the image size is 320 ×.
240. An image to be input will be described by taking an example in which luminance and chromaticity are separated.

A control device (not shown) sequentially reads the frames to be coded one by one, instructs the mode of the frame to be the intra-frame coding mode, the interlaced mode, or the bidirectional motion compensation mode, and operates the encoder. Is controlled.

Prior to encoding, the control device also initializes each device. The threshold value of the comparator is set to 600.

When the frame to be encoded is in the intra-frame encoding mode, the encoding is performed as follows. Terminal 12
The original image is input from 3 and stored in the frame memory 100. The coding mode device determiner 112 fixes the intra-block coding mode in which the intra-block coding is performed until the coding of this frame is finished, and the counter 120 does not operate in this frame. Further, the selector 104 has a buffer 126.
Leave empty. According to an instruction from the control device, original blocks are cut out from the frame memory 100 in the main scanning direction and input to the selector 104. Since the selector 104 always selects the output of the frame memory 100 in this frame mode, the extracted original block is delayed by the delay device 113.
It is input to the selector 116 via. Since the coding mode determiner 112 continues to output the intra-block coding mode, the output of the delay device 113 is always selected in this frame. The encoder 119 encodes the image size and that the mode of the present frame is the intra-frame encoding mode at the beginning of the frame and outputs the encoded image size to the buffer 126. Following this, the encoder 119 performs a discrete cosine transform for each original block, quantizes the obtained transform coefficient and assigns a code, stores the code in the buffer 126, and at the same time decodes the decoder 118.
Output to. The buffer 126 waits for an instruction from the control device and outputs the signal from the terminal 125 to the outside after finishing the coding for all the blocks of the frame. On the decoding side, first, the frame mode and the image size are received and decoded.
Decoding is continued and the quantization result of the discrete cosine transform is set to 8
As soon as all the × 8 coefficients are accumulated, inverse quantization and inverse discrete cosine transform are performed to obtain decoded pixel values. The playback position of the block is naturally determined from the image size, and after the last block is decoded, it waits until the next frame mode comes. The same thing is performed by the decoder 118, and the decoded image is stored in the frame memory 102. After that, the control device causes the buffer 126 to start transmission from the terminal 125. When the transmission is completed, the buffer 126 informs the control device of the completion of the transmission. After detecting the end of transmission of the control device, the encoding of the next frame is started.

When the input frame is in the bidirectional motion compensation mode, the frame input from the terminal 123 is stored in the frame memory 101, and the next frame (interlaced frame or intra-frame coded frame) is input. The data is stored in the memory 101, the interlaced frame or the intra-frame encoded frame is encoded first, and then the bidirectional motion compensation frame is processed.

When the frame to be encoded is in the interlace mode, the encoding is performed as follows. Before the processing of a frame is started, if the immediately previous frame is the intraframe coding mode, the intraframe coding mode of the coding mode determiner 112 is released. Further, the contents of the buffer 126 are cleared, and the counter 120 resets the count value to 0. Further, the selector 104 is made to select the output from the frame memory 101, and the buffer 126 is emptied. After that, the original image is input from the terminal 123 to the data of the next frame, and the frame memory 1
Stored in 00.

According to an instruction from the control device, original blocks are cut out from the frame memory 100 in the order of the main scanning direction, passed through the selector 104, and then the motion vector detector 105.
To enter. The motion vector detector 105 inputs the original block data and the original block position in the image of the frame memory 102 (the immediately preceding coded intra-frame coded frame or the coded interlaced frame) (this is the number of blocks). Uniquely determined by counting), the motion vector is detected by full search for the area of ± 31 pixels in the main scanning and sub-scanning directions with the position of the original block as the center. The detected motion vector is input to the motion compensator 107. The motion compensator 107 cuts out the prediction block data from the frame memory 102 according to this motion vector. Frame memory 10 at the same time
The original block is input from 0 through the selector 104, a difference is calculated for each pixel from the prediction block, and a prediction error block composed of this difference value is obtained and output. After that, the original block and the prediction error block are delayed by the delay units 113 and 114, respectively.
Stored in. The data of the original block is the block evaluator 1
09, and first, the sum of squares of pixel values in the block S ₀
Is calculated. Similarly, the data of the prediction error block is input to the block evaluator 110, and the square sum S _E of the intra-block error value is calculated. These sums of squares S ₀ and S _E are input to the coding mode determiner 112. Coding mode determiner 1
12 determines the coding mode of this block from the magnitude relation of the following equation of the sum of squares.

IfS _E > S ₀ : The intra-block coding mode is the coding mode ifS _E ≦ S ₀ : The coding mode is the motion guarantee mode

This coding mode is input to 120 in the counter scene change detector 127 and counts the number of coding blocks that were in the intra-block coding mode. At the same time, the coding mode is also supplied to the selectors 116 and 117,
Delay device 1 if the coding mode is intra-block coding
In the motion compensation mode, the original block data which is the output of 13 is selected and the prediction error block which is the output of the delay device 114 is selected and output. The encoder 119 first encodes that the mode of this frame is the interlaced frame and the image size, and outputs the encoded image size to the buffer 126. Then, each block is encoded. First, the coding mode of a block is coded, and subsequently, when the coding mode is the intra-block coding mode, the original block is subjected to discrete cosine transform, the obtained transform coefficient is quantized, and the code is assigned. If the encoding mode is motion compensation mode, first,
The motion vector is encoded, the obtained prediction error block is then subjected to discrete cosine transform, and the obtained transform coefficient is quantized to assign a code. These codes are stored in the buffer 126 and simultaneously output to the decoder 118.
The decoder 118 first receives and decodes the frame mode and the image size as in the intraframe coding mode.
Further, the coding mode is decoded for each block, and if the coding mode is the intraframe coding mode, the inverse quantization and the inverse discrete cosine will be applied as soon as all 8 × 8 coefficients of the quantization result of the discrete cosine transform are accumulated. Transformation is performed to obtain a decoded pixel value.
If the encoding mode is the motion compensation mode, the subsequent motion vector value is decoded. The pixel block at the position moved by the motion vector from the previously decoded image is read out. The quantization result of the discrete cosine transform coefficient following the motion vector is decoded, and as soon as all 8 × 8 coefficients of the decoded result are accumulated, inverse quantization and inverse discrete cosine transform are applied,
A decoding error block can be obtained. This error block is added to the previously decoded pixel block to obtain a decoded pixel value. The reproduction position of the block is naturally determined from the image size, and after the last block is decoded, it waits until the encoded data of the next frame comes. The same applies to the decoder 11
8 and stores the decoded image in the frame memory 103.

After performing the encoding process on all the blocks, the counter 12 in the scene change detector 127
For 0, the count value of the intra-block coding mode is input to the comparator 121 and compared with the threshold value set by the threshold value determiner 122. If the number of blocks that were in intra-block coding mode is greater than or equal to the threshold, then terminal 124
Informs the control unit of the detection of a scene change from. Upon detection of the scene change, the control device stops the entire operation, changes the mode of the frame to be encoded from the interlace mode to the intraframe encoding mode, and re-encodes the frame. If the count value is lower than the threshold value, the controller 126 holds the contents of the frame until the coding of the next bidirectional motion compensation frame is completed.
In parallel with this, the count value output from the counter 120 is input to the threshold value determiner, and twice the count value is updated as a new threshold value for the interlaced mode.

When the frame to be encoded is in the bidirectional motion compensation mode, the encoding is performed as follows. Before the frame processing is started, if the immediately following frame is in the intraframe coding mode, the intraframe coding mode of the coding mode determiner 112 is released. Further, the counter 120 resets the count value to 0. Further, the selector 104 is made to select the output from the frame memory 101, and the buffer 126 is emptied.

It is stored in the frame memory 101 in the bidirectional motion compensation mode according to an instruction from the control device.
The original block is cut out from the frame memory 101 in the order of the main scanning direction, and is input to the motion vector detectors 105 and 106 via the selector 104. Motion vector detector 1
05 is the input original block data and the frame memory 1
Center the original block position from the original block position (this is uniquely determined by counting the number of blocks) in the 02 image (previously encoded interlaced frame or encoded intra-frame encoded frame) As a result, a motion vector is detected by full search for an area of ± 31 pixels in the main scanning and sub scanning directions. Similarly, the motion vector detector 106 detects the original block position from the input original block data and the original block position in the image of the frame memory 103 (immediately after the encoded interlaced frame or the encoded intra-frame encoded frame). A motion vector is detected by full search with respect to a region of ± 31 pixels in the main scanning direction and the sub scanning direction centering on the position. The detected motion vector is input to the motion compensators 107 and 108, respectively. Here, referring to the immediately preceding frame is called forward, and referring to the immediately following frame is called backward. The motion compensator 107 cuts out the forward prediction block data from the frame memory 102 according to the forward motion vector. At the same time the frame memory 101
Then, the original block is input via the selector 104, the difference is calculated for each pixel from the forward prediction block, and the forward prediction error block composed of this difference value is obtained and output. Similarly, the motion compensator 108 also cuts out the backward prediction block data from the frame memory 103 according to the backward motion vector. At the same time, the original block is input from the frame memory 101 via the selector 104, the difference between the backward prediction block and each pixel is calculated, and the backward prediction error block consisting of this difference value is obtained and output. After that, the original block, the forward prediction error block, and the backward prediction error block are stored in the delay units 113, 114, and 115, respectively.
The data of the original block is input to the block evaluator 109, and first, the sum of squares S ₀ of the pixel values in the block is calculated.
Similarly, the data of the forward prediction error block is input to the block evaluator 110, the sum of squares S _F of the intra-block error values is calculated, and the data of the backward prediction error block is input to the block evaluator 111 to calculate the intra-block error value. Sum of squares S _B
Is calculated. These sums of squares S ₀ , S _F and S _B are input to the coding mode determiner 112. The coding mode determiner 112 determines the coding mode of this block from the magnitude relationship of the following sum of square sums.

Ifmax (S _F , S _B )> S ₀ : The coding mode is an intra-block coding mode ifmax (S _F , S _B ) ≦ S ₀ ifS _F > S _B : The coding mode is a backward motion compensation mode ifS _F ≤ S _B : Encoding mode is forward motion compensation mode

This coding mode is input to the counter 120 in the scene change detector 127 and counts for each coding mode. At the same time, the encoding mode is selector 116,
When the coding mode is intra-block coding, the original block data which is the output of the delay device 113 is output to the output block 117, and the prediction error block which is the output of the delay device 114 is output in the backward motion compensation mode. In the motion compensation mode, the prediction error block of the delay device 115 is set to the selector 11
6 are selected and output. The encoder 119 first encodes that the mode of this frame is the bidirectional motion compensation frame and the image size, and outputs the encoded image size to the buffer 126. The following code the coding mode of the block to be coded for each block, and subsequently, if the coding mode of this block is the intra-block coding mode, the original block is subjected to the discrete cosine change to obtain Quantize the transform coefficients and assign codes. If the coding mode is the forward / backward motion compensation mode, first, the motion vector is coded, then the obtained prediction error block is subjected to discrete cosine transform, and the obtained transform coefficient is quantized and coded. assign. These codes are output to the decoder 118 at the same time as being stored in the buffer 126. The decoder 118 first receives and decodes the frame mode and the image size, as in the interlaced frame. Further, the coding mode is decoded for each block, and if the coding mode is the intra-frame coding mode, 8 × of the quantization result of the discrete cosine transform is obtained.
As soon as all the coefficients of 8 are accumulated, inverse quantization and inverse discrete cosine transform are performed to obtain a decoded pixel value. If the encoding mode is the motion compensation mode, the subsequent motion vector value is decoded. The pixel block at the position moved by the motion vector from the already decoded previous or subsequent image is read out. Decoding the quantization result of the discrete cosine transform coefficient following the motion vector, and as soon as all 8 × 8 coefficients of the decoding result are accumulated,
Inverse quantization and inverse discrete cosine transform are applied to obtain a decoding error block. This error block is added to the previously decoded pixel block to obtain a decoded pixel value. The playback position of a block is naturally determined from the image size, and after the last block is decoded, it waits for the next frame.
A similar operation is performed by the decoder 118, and the frame memory 1
The decoded image is stored in 03.

After performing the encoding process for all blocks, the counter 12 in the scene change detector 127
For 0, the count value of each coding mode is input to the comparator 121 and compared with each threshold set by the threshold determiner 122. The count value of each coding mode is as follows.

The count value of the intra-block coding mode is B _{I The} count value of the forward motion compensation mode is B _{T The} count value of the backward motion compensation mode is B _B

Further, the threshold of the intra-block coding mode is T
Let h ₁ . If the number of blocks B ₁ that were in intra-block coding mode is greater than or equal to the threshold Th _1, it informs the controller of the detection of a scene change from terminal 124 to the previous frame. In parallel with this,
A difference value B _S between the count value B _{T in the} forward motion compensation mode and the count value B _{B in} the backward motion compensation mode is calculated according to the following equation.

B _S = B _F −B _B

The difference value B _S is compared with the threshold value Th _S, and if B _S > Th _S , the control device is notified of the detection of the scene change between the terminal 124 and the next frame. When the controller receives a scene change from the previous frame, the whole operation is stopped, the mode of the frame to be encoded is changed from the bidirectional motion compensation mode to the intra-frame encoding mode, and the buffer 126 is set. Empty it and re-encode the frame. If the control unit detects a scene change between the previous frame and the scene change is not detected during the encoding of the next frame, the whole operation is stopped and the mode of the frame to be encoded is changed to both directions. Change from motion compensation mode to intraframe coding mode,
After the buffer 126 is emptied and the next interlaced frame is re-encoded in the intra-frame encoding mode, the frame immediately before (the scene change is detected) is encoded in the bidirectional motion compensation mode.

Otherwise, the controller will use the buffer 12
6 outputs the contents of the frame from the terminal 125 to the outside.
In parallel with this, the count value output from the counter 120 is input to the threshold value determiner, and twice the count value is updated as a new threshold value for the bidirectional motion compensation mode.

(Other Embodiments) The block size, encoder coding system, motion compensation, motion vector search, intra-block evaluation system, frame memory configuration, etc. are not limited to these. For example, in a configuration in which a plurality of bidirectional motion compensation frames continue, in addition to comparing the difference value B _S with the threshold Th _S depending on the position of the frame, a scene change from the previous frame is performed by −B _S > Th _S. It can also be detected.

[0038]

EFFECTS OF THE INVENTION A scene change can be detected with a simple structure without providing a means for detecting a scene change other than encoding, and a frame in which a scene change is detected is intra-block encoded to deteriorate a decoded image. Can be suppressed. Also, by dynamically changing the threshold of scene change detection while encoding,
It can detect even moving images that move rapidly.

[Brief description of drawings]

FIG. 1 is a block diagram of a moving picture coding apparatus embodying the present invention.

FIG. 2 is a block diagram of a moving picture coding apparatus embodying the present invention.

FIG. 3 is a diagram showing an arrangement of encoded frames in the first embodiment.

FIG. 4 is a diagram showing an arrangement of encoded frames in the second embodiment.

FIG. 5 is a block diagram showing a conventional example of a moving image encoding device.

[Explanation of symbols]

1, 2, 15, 100, 101, 102, 103, 50
1,502 Frame memory 3,105,106,503 Motion vector detector 4,107,108,504 Motion compensator 5,6,109,110,111,505,506 Block evaluator 7,112,507 Coding mode Determinator 8, 9, 113, 114, 115, 508, 509 Delay device 10, 104, 116, 117, 510 Selector 11, 120 Counter 12,121 Comparator 13, 119, 511 Encoder 14, 118, 512 Decoder 16, 17, 18, 19, 123, 124, 125, 5
14,515 Terminal 20,126 Buffer 21,127 Scene change detector 122 Threshold value determiner

Claims (11)

[Claims] 1. A method for classifying and coding a moving image composed of a plurality of frames into a frame for intra-frame coding and a frame for motion-compensating and coding, and detecting a scene change. , A moving picture coding method characterized in that intra-frame coding is performed again for a scene change frame. 2. A moving image in which a frame to be encoded is divided into a plurality of blocks, a motion vector is calculated with reference to an already encoded frame corresponding to each block, a prediction block is obtained, and a prediction error is encoded. A moving picture coding apparatus characterized in that, in the coding apparatus, a motion vector used at the time of coding is counted for each reference frame and a scene change is detected by the counted value. 3. A moving image code for dividing a frame to be encoded into a plurality of blocks, calculating a motion vector by referring to an already encoded frame corresponding to each block, obtaining a prediction block and encoding a prediction error. In the encoding device, a moving image encoding device is characterized in that a block to be encoded in the block which completes the encoding in the encoding is counted for each frame, and a scene change is detected by the counted value. . 4. A scene change is detected by comparing a count value of each motion vector used for encoding for each reference frame with a predetermined threshold value.
The moving picture encoding device described. 5. The moving image according to claim 2, wherein a scene change is detected by comparing a count value of each motion vector used for encoding for each reference frame with a dynamically determined threshold value. Image coding device. 6. A scene change is detected by comparing a count value of each motion vector used for encoding for each reference frame with a predetermined threshold value.
The moving picture encoding device described. 7. The moving image according to claim 3, wherein a scene change is detected by comparing a count value of each motion vector used in encoding for each reference frame with a dynamically determined threshold value. Image coding device. 8. A method of dividing a frame to be encoded into a plurality of blocks, calculating a motion vector with reference to an already encoded frame corresponding to each block, obtaining a prediction block, and encoding a prediction error. In a moving picture coding apparatus for intra-block coding the block when the prediction error is large, a means for counting a motion vector used at the time of coding for each frame to be referenced, and comparing the count value with a threshold value. And a means for judging the presence or absence of a scene change from the result of the comparison. 9. The moving picture coding apparatus according to claim 8, further comprising means for updating, for each frame, a threshold value from a count value of the number of blocks for which intra-block coding of a frame has already been coded. 10. A moving image in which a frame to be encoded is divided into a plurality of blocks, a motion vector is calculated with reference to an already encoded frame corresponding to each block, a prediction block is obtained, and a prediction error is encoded. In image coding, means for counting the motion vector used at the time of coding for each frame to be referred to, means for comparing the count value with a threshold value, and means for determining the presence or absence of a scene change from the result of the comparison are provided. A moving picture coding apparatus having. 11. The moving picture coding apparatus according to claim 10, further comprising means for updating the threshold value for each frame from the count value of the motion vector of the already coded frame.