JPH0530498A - Moving picture data compressing device - Google Patents

Abstract

PURPOSE:To discriminate whether an input block is effective or not with a high precision and to accurately detect a motion vector. CONSTITUTION:Preceding input raw moving picture data corresponding to one frame which is not processed is stored in a motion compensation inter-frame forecast device 14. The difference between raw moving picture data of an arbitrary block out of stored raw moving picture data corresponding to one frame and an input block is calculated to output a motion compensation forecast value. A second subtractor 15 calculates the difference between newly inputted moving picture data corresponding to one block and the motion compensation forecast value to output a motion compensation forecast error. An effective/ ineffective discriminator 16 discriminates whether the picture is moved or not based on this forecast error to judge the input block to be effective or ineffective.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving image data compression apparatus for reducing the information amount of moving image data.

[0002]

2. Description of the Related Art This kind of moving image data compression is used in videophones, videoconferences, etc., and an orthogonal transform coding system is in the limelight as the data compression system. As a conventional moving image data compression apparatus using this method, the Institute of Television Engineers, Vol. 43, No. 10 (1989)
2 has the configuration shown in FIG.

The video signal is input to the subtractor 1 for each block, and the difference from the prediction value output from the motion-compensated interframe prediction means 3 via the in-loop filter 2 is calculated. This calculation result corresponds to the motion-compensated inter-frame prediction error, and the orthogonal transform means 4 calculates the orthogonal transform coefficient for each block of the input moving image data. The transform coefficient is quantized by the quantizing means 5, data is compressed, and is transmitted. Further, the quantized transform coefficient is decoded into the original prediction error by the inverse quantizing unit 6 and the inverse orthogonal transforming unit 7. Further, the decoded prediction error is added to the adder 8
At, it is converted into a locally decoded output and given to the motion-compensated interframe prediction means 3. The motion-compensated inter-frame prediction means 3 performs motion-compensated inter-frame prediction based on the input local decoding output. Further, the validity / invalidity determining means 9 inputs the prediction error output from the subtracter 1, determines whether or not there is a motion in the input image based on the prediction error, and determines the validity of the input block. When there is no motion, the input moving image data is set as an invalid block, and an identification signal indicating the invalid block is transmitted without performing orthogonal transformation.

[0004]

However, in the above-described conventional moving picture data compression apparatus, the following problems occur when encoding at a low rate of about 64 Kbps.
That is, when low-rate encoding is performed, it is necessary to increase the quantization level in the quantizing means 5 to increase the compression efficiency. Therefore, image distortion such as block distortion occurs in the transmission data image decoded by the inverse quantization unit 6 and the inverse orthogonal transform unit 7. Therefore, the distortion also affects the prediction value predicted by the motion-compensated interframe predictor 3 based on the decoded image including the block distortion, and the subtractor 1 calculates the prediction value from the input video signal. The prediction error signal will be inaccurate. If the prediction error is inaccurate, valid / based on this prediction error
The determination by the invalidation determination unit 9 may be incorrect. That is, even if the block of interest exists in the still image area, it is determined that the image has moved due to the influence of distortion, and it may be erroneously determined to be valid. In addition, since the motion vector detection in the motion compensation inter-frame prediction means 3 is also performed based on the decoded image affected by the image distortion, an incorrect motion vector may be detected.

As described above, in the conventional moving image data compression apparatus, an invalid block that is not valid may be unnecessarily transmitted, or an input image may be encoded with an incorrect motion vector. As a result, the amount of generated codes increases, so that the quantization level must be further increased, resulting in a vicious circle in which block distortion occurs more and more.

[0006]

SUMMARY OF THE INVENTION The present invention has been made to solve such a problem, and a second moving image storing live moving image data which has not been processed for one frame input last time is stored. The storage means and the difference between the live moving image data at the frame position corresponding to the newly input moving image data among the live moving image data stored in the second moving image storage means and the newly input moving image data this time. A second subtraction means for calculating the above, and a judging means for judging whether or not a motion has occurred in the image based on the calculation result and judging whether the moving image data newly input this time is valid or invalid. It is characterized by.

Further, the second moving image storage means stores the live moving image data for one block at an arbitrary frame position out of the previously input one frame of live moving image data stored and the newly input moving image for this time. The difference from the data is calculated, and one block of live image data at the frame position where the difference is minimized is output as the motion compensation prediction value, and the second subtraction unit newly inputs this motion compensation prediction value and this time. It is characterized in that the motion compensation prediction error is output by calculating the difference from the moving image data for one block.

The second moving image storage means uses the amount of motion between the live moving image data for one block at the frame position where the difference is the minimum and the moving image data newly input this time as a motion vector. To the moving image storage means of
Of the stored moving image data outputs moving image data for one block at the frame position indicated by the motion vector, and the adding means
This moving image data output from the first moving image storage means and the decoded prediction error signal output from the inverse orthogonal transform processing means are added, and the newly input moving image data is decoded to obtain the first moving image data. It is characterized in that it is stored in the moving image storage means.

[0009]

The valid / invalid determination of the newly input moving image data in the determining means is performed in comparison with the previously input live moving image data which has not been processed.

Further, since the validity / invalidity judgment is performed in comparison with the motion compensation prediction value and the new input block, the probability that the judgment result becomes invalid is increased.

Furthermore, the motion-compensated prediction is performed based on the live-moving image data, and the motion vector detection becomes accurate.

[0012]

1 is a block diagram showing the configuration of a moving image data compression apparatus according to an embodiment of the present invention.

A video signal, which is moving image data, is input to the first subtractor 11. This video signal is image data for one frame (one screen) divided into blocks of 8 × 8 bits, and moving image data is input to the subtractor 11 for each block. The motion-compensated inter-frame predictor 14 is composed of a frame memory, and the motion-compensated inter-frame predictor 14 includes one frame of live moving image data input last time, that is, raw data that has not been processed yet. Moving image data is stored. This motion-compensated inter-frame predictor 14 uses the above-mentioned 8
The difference between the × 8-bit block data and the moving image data of an arbitrary 8 × 8-bit block of the stored moving image data of one frame is calculated, and the block data with the smallest difference is calculated. It outputs to the subtracter 15 of 2 as a motion compensation prediction value. By calculating this difference, the position of the current input block in the previous one frame of the moving image data is determined, and the motion-compensated inter-frame predictor 1
4 outputs to the frame memory 12 the amount of motion from the previous block position to the current input block position as a motion vector. At the same time, the motion compensation interframe predictor 14 sends this motion vector to the receiving side.

The second subtractor 15 compares the moving image data for one block newly input this time with the motion compensation prediction value output from the motion compensation inter-frame predictor 14, and calculates the difference. . This calculation result corresponds to the motion compensation prediction error and is given to the valid / invalid determination unit 16. Based on this motion compensation prediction error, the valid / invalid determiner 16 determines whether the current input block is in the still image area or the moving image area. That is, comparing the given motion-compensated prediction error with a predetermined threshold,
When the prediction error exceeds the threshold value, it is determined that the input block is in the moving image area. If the threshold is not exceeded, it is determined that the input block is in the still image area and the input block is invalidated. Then, this determination result is sent to the receiving side.

Further, the frame memory 12 stores the previously input one frame of moving image data which has been quantized and decoded. Of the decoded previous one frame of moving image data, the motion-compensated inter-frame predictor 1
4, the moving image data of the block designated by the motion vector output from No. 4 is extracted, and the in-loop filter 13
Is input as a predicted value to the first subtractor 11 via. The subtractor 11 receives the newly input 8 while the motion compensation is being performed by the motion compensation inter-frame predictor 14.
× 8 bit block data is stored. And
The difference between each pixel of the image data extracted from the frame memory 12 based on the motion vector and the video signal newly input and stored this time is calculated, and the calculation result is output as a prediction error signal.

This prediction error output from the subtractor 11 is given to the switch 17a. The switching of the switch 17a and the switch 17b is controlled by the scene change detector 18. For example, if new image data is input, or if the scene changes completely, the image data newly input this time has no correlation with the previously input image data, and image continuity is lost. ing. Therefore, the scene change detector 18 switches the connection of each of the switching devices 17a and 17b to the terminal located on the upper side of the drawing, and the subtractor 1
The output of 1 is ignored and the inter-frame prediction is not performed, and the deterioration of the coding efficiency is suppressed. When the images have continuity, the scene change detector 18 switches the connection of each of the switching devices 17a and 17b to the terminal located on the lower side of the drawing, and gives the output of the subtractor 11 to the DCT converter 19. Control. With this switching control, the following inter-frame prediction is performed.

The DCT converter 19 is a two-dimensional DCT (Discrete Cos) for the prediction error signal input from the subtractor 11.
ine Transform: Discrete cosine transform) is applied to remove spatial correlation of prediction error for each 8 × 8 block size, and energy is concentrated on some transform coefficients. That is, in a signal having a strong correlation, the signal power is concentratedly distributed on a specific frequency component when viewed on the frequency axis. Therefore, the amount of information can be reduced by coding only the coefficients of the components in which the signal power is concentrated and distributed by this DCT conversion. The quantizer (Q) 20 quantizes the transform coefficient output from the DCT converter 19 into the number of levels according to the energy, performs data compression, and sends the quantized level to the receiving side.

The inverse quantizer (Q ^-1 ) 21 is a quantizer 20.
The quantization level output from is dequantized and returned to the original transform coefficient before being quantized. The inverse DCT (IDCT) converter 22 further performs inverse DCT conversion on the transform coefficient dequantized by the inverse quantizer 21, and decodes it into the original prediction error signal output from the subtractor 11. Adder 23
Of the previously decoded moving image data stored in the frame memory 12, the block data designated by the motion vector is input via the switch 17b.
Then, the prediction error signal decoded by the IDCT converter 22 and the previous image data input from the switch 17b are added, and the block moving image data newly input this time is locally decoded. The decoded moving image data is replaced with the previous corresponding block data stored in the frame memory 12. By performing this inverse conversion on all input block data for one screen,
The stored contents of the frame memory 12 are converted into one frame of moving image data input this time and used as a predicted value of moving image data input next time.

According to this embodiment as described above, the motion prediction in the motion-compensated inter-frame predictor 14 is not the previous moving image data once quantized and decoded in the conventional data compression apparatus, but the previous raw image data. Is performed based on the moving image data of. Therefore, it is not affected by the conventional block distortion that occurs in the decoded screen, and the motion prediction is accurately performed. Therefore, it is not necessary to detect an erroneous motion vector as in the conventional case. Further, since the motion prediction value output from the motion compensation inter-frame predictor 14 is accurate, the prediction error output from the second subtractor 15 is also accurate, and as a result, the valid / invalid determiner 16 The valid / invalid judgment of the input block in is accurate. Therefore, unlike the conventional case, the block in the still image area is not mistakenly determined to be a valid block,
The amount of code to be encoded does not unnecessarily increase.

Moreover, since the target block of the subtractor 15 which takes the difference from the new input block is the block whose motion is predicted by the motion-compensated interframe predictor 14, the probability that the difference becomes zero becomes high. That is, in the previously input one frame of moving image data, an arbitrary block having a pixel configuration closest to the new input block is extracted by the motion compensation inter-frame predictor 14 and given to the subtractor 15. is there. If the probability that the difference result in the subtractor 15 becomes zero becomes high, the judgment result in the valid / invalid judgment unit 16 naturally becomes high in probability, the amount of information to be transmitted is reduced, and the data compression becomes easy. .

In the above description of the embodiment, the motion compensation inter-frame predictor 14 is used for motion compensation, but it is not always necessary to perform this motion compensation. When motion-compensated prediction is not performed, the same effect as in the above-described embodiment can be obtained by placing a simple frame memory instead of the motion-compensated inter-frame predictor 14. In other words, the previously input raw moving image data is stored in this frame memory, and the difference between the block data newly input in the subtractor 15 and this raw moving image data is taken to obtain a conventional decoding screen. This is because it is not affected by the generated block distortion.

[0022]

As described above, according to the present invention, the validity / invalidity determination of the moving image data in the determiner is performed in comparison with the raw moving image data which has not been processed. Therefore, the validity / invalidity of the input block is determined by using an image without distortion, and the determination result becomes accurate. As a result, it is erroneously determined to be valid and the unnecessary block is transmitted as in the conventional case. Things will disappear.

Since the validity / invalidity determination is performed in comparison with the motion compensation prediction value and the new input block, the determination result is more likely to be invalid. Therefore,
The amount of code to be encoded is reduced, and information compression becomes easier.

Furthermore, since the motion-compensated prediction is performed based on the raw moving image data, the motion vector detection is accurately performed. Therefore, encoding is not performed based on an incorrect motion vector, and the information compression rate is improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a moving image data compression apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a conventional moving image data compression device.

[Explanation of symbols]

11 ... First subtractor 12 ... Frame memory (first moving image storage means) 13 ... In-loop filter 14 ... Motion compensation interframe predictor (second moving image storage means) 15 ... Second subtractor 16 ... Effective / invalid determiner 17a, b ... Switch 18 ... Scene change detector 19 ... DCT converter 20 ... Quantizer 21 ... Inverse quantizer 22 ... Inverse DCT converter 23 ... Adder

Claims (3)

[Claims] 1. A first moving image storage means for storing the previously input one frame of decoded moving image data, and the moving image data stored in the first moving image storage means this time. A first operation of calculating a difference between moving image data at a frame position corresponding to newly input moving image data for one block and the moving image data newly input this time, and outputting the difference value as a prediction error signal. Subtraction means, orthogonal transformation processing means for removing the spatial correlation of the prediction error signal and concentrating the energy in a part of the transform coefficients, and quantizing the transform coefficients into the number of levels according to the energy to obtain the data. Quantizing means for performing compression and transmitting the quantized level, dequantizing means for dequantizing the quantized level to return it to the original transform coefficient, and inverse quantizing for the inverse quantized transform coefficient. An inverse orthogonal transform processing unit that performs a transform process to decode the original prediction error signal, and the decoded prediction error signal and the moving image data stored in the first moving image storage unit are added to this time to newly add An adding means for decoding the moving image data input to the first moving image storage means and storing the decoded moving image data in the first moving image storage means; and a second moving image storing the previously input one frame of unprocessed live moving image data. Storage means, and of the live moving image data stored in the second moving image storing means, live moving image data at a frame position corresponding to the moving image data newly input this time, and the moving image newly input this time. Second subtraction means for calculating the difference from the image data, and based on the result of the calculation, it is judged whether or not motion has occurred in the image to judge whether the newly input moving picture data is valid or invalid. Moving image data compression apparatus characterized by comprising a determining means. 2. The second moving image storage means stores 1 block of live moving image data at an arbitrary frame position among the previously input 1 frame of live moving image data stored therein.
The difference from the newly input moving image data is calculated, and one block of live moving image data at the frame position where the difference is minimized is output as a motion compensation prediction value. Calculate the difference between the predicted value and the moving image data for one block newly input this time,
The moving image data compression apparatus according to claim 1, wherein the difference value is output as a motion compensation prediction error. 3. The second moving image storage means stores live moving image data for one block at a frame position where the difference is minimized.
The motion amount with respect to the moving image data newly input this time is output as a motion vector to the first moving image storage means, and the first moving image storage means uses this moving image data among the stored moving image data. The moving image data for one block at the frame position indicated by the vector is output, and the adding unit decodes the moving image data output from the first moving image storage unit and the inverse orthogonal transform processing unit. 3. The moving image data compression apparatus according to claim 2, wherein the moving image data newly input this time is decoded by adding the generated prediction error signal and stored in the first moving image storage means.