JPH0783479B2 - Method and apparatus for encoding moving image signal - Google Patents

Description

TECHNICAL FIELD The present invention relates to a digital band compression technique for a moving image using an encoding method that utilizes correlation between screens.

(Prior Art) Conventionally, as an encoding method using correlation between screens, for example, an interframe predictive encoding method using motion compensation is known (Ninomiya et al., "Motion Compensation Interframe Encoding Method"). Academic theory (B), J63-B, 11, pp.1140-1147, Sho 51-11, "Reference 1").

This method divides the screen into small blocks, calculates the block with the highest correlation in the stored previous screen for each block, and calculates the difference in position (motion vector) between the corresponding blocks and this corresponding block. This is a method of transmitting a difference in amplitude value (motion compensation prediction error) of pixels at the same spatial position between the blocks.

(Problems to be Solved by the Invention) However, in the above-described conventional technique, when the contents of the screen change significantly such as a scene change, the correlation between frames is significantly reduced, and a large amount of prediction error to be transmitted occurs. Occurrence and transmission at a constant transmission rate, rough coding mode such as quantization of rough characteristics, field dropping, frame dropping, and worst case coding stop in the middle of the screen is applied. I had to do it. As a result, when encoding is stopped, an area that has just been encoded within the screen (new screen) and an area that has not yet been encoded (previous screen) freezes in a mixed state. Since the freshly coded area is also often coded in the rough coding mode, there is a drawback that it causes a very noticeable deterioration such as a large deterioration in image quality.

(Means for Solving the Problems) According to the present invention, the encoding of the moving image signal in which the input moving image signal temporally changes from the Nth screen to the (N + 1) th, (N + 2) ... It is determined whether or not the N + 1 screen has changed significantly from the Nth screen, and when the determination is determined to have changed significantly, a screen in which a certain area of the Nth screen is rewritten with the N + 1th screen is set as an encoding target. The moving image signal is encoded by using the correlation with the Nth screen, which is the screen of the previous time, and then the certain region is gradually widened with the passage of time, and the widened region is sequentially arranged in the N + 2th, N + 3 ... A code of a moving image signal characterized in that the image signal rewritten on the (N + M) th screen is set as a moving image signal to be encoded, and is encoded by using the correlation with the image signal to be encoded in the previous time. Got the method That.

Further, according to the present invention, a determination unit that determines whether the input moving image signal has significantly changed the contents of the screen from the moving image signal of the previous time, and a synchronization signal is generated from the input moving image signal. A synchronizing signal generating means; a delaying means for delaying the input moving image signal by a predetermined time; a first counting means for synchronizing an address of the screen in one frame time of the screen in synchronization with the synchronizing signal; When it is determined that the content of the screen has changed significantly by the output of the determination means, the address for designating a certain area of the screen for each predetermined screen is gradually expanded in synchronization with the synchronization signal. The second counting means for outputting, the comparing means for comparing the addresses of the first counting means and the second counting means, and outputting the control signal for controlling the rewriting of the certain area, and the control signal The above A storage unit that rewrites and stores a certain area of the screen of the previous time on the screen of the moving image signal output from the extending unit, and outputs this as the moving image signal of the encoding target, and the moving image signal of the encoding target. A coding device for a moving picture signal, comprising: a coding means for coding by using a correlation with a moving picture signal to be coded at a previous time.

(Operation) First, in the present invention, it is necessary to detect a case in which the content changes greatly between screens. As the method, a method of detecting a screen change (scene change) using a motion compensation prediction error value will be described.

The motion compensation prediction error can be calculated, for example, by using the motion vector detected by the block matching method described in Document 1 above. What is the block matching method?
As shown in FIG. 2, the past screen F ′ is stored in the frame memory, and the block A ′ having the highest correlation with the block A in the current frame F is obtained in the stored past screen. The displacement of the position on the screen is used as a motion vector. The motion compensation prediction error corresponds to the difference value for each pixel at the spatially same position in the two blocks associated by the motion vector. As the evaluation standard of the height of the correlation, the square of the difference value between the corresponding pixels or the minimum sum of absolute values in the block is used.

The present invention and the method of obtaining the motion compensation prediction error are not directly related,
The motion compensation prediction error may be obtained by a method other than the above.

FIG. 3 is a diagram showing how the dispersion value of the motion compensation prediction error and the inter-frame difference value changes with respect to the speed of movement between frames (the speed of panning). The variance value of the inter-frame difference value is expressed by an equation, and increases monotonically as the movement becomes faster. On the other hand, the variance value of the motion compensation prediction error value is expressed by an expression, and when the motion is an integer multiple of the pixel interval, the motion compensation prediction is correct, so the prediction error is zero at this time as shown in the figure. Become.

Here, ▲ σ ² _FF ▼ and ▲ σ ² _MC ▼ represent the variance values of the frame difference value and the motion compensation prediction error value, respectively, and ▲ σ ² _O ▼
Is the average power of the image, and α is a parameter corresponding to the complexity of the screen.

▲ σ ² _FF ▼ = 2 ▲ σ ² _O ▼ ・ {1-exp (−α |｜) ▲ σ ² _MC ▼ = 2 ▲ σ ² _O ▼ ・ {1-exp (−α ・ | −
[] | [V] represents an integer part of the meanwhile, in the case of an image such as a scene change where there is no correlation between frames, the inter-frame difference value and the variance value of the motion compensation prediction error are both relative to the inter-frame difference value in FIG. The value is close to the upper limit of variance. Therefore, when the inter-frame difference value is used, it is impossible to distinguish between a scene change and a fast-moving image, but by using the motion-compensated prediction error to compare with the determination threshold as shown in FIG. Only scene changes can be detected accurately.

The detection method in the case where the content changes greatly between the screens including the sea change has been described above, but the detection method itself is not limited to the above method in the present invention. In the following, assuming that the contents have been largely changed between the screens has already been demanded, a method of gradually expanding the area for rewriting the contents of the input frame memory (hereinafter referred to as curtain control) will be described.

In the Nth screen time of FIG. 1, the screen N is coded.
Then, it is assumed that the contents of the screen are largely changed from the Nth screen time to the (N + 1) th screen time. At this time, it is assumed that a scene change is detected when the screen N + 1 is encoded up to the L-th line by the above-mentioned scene change detection method or the like. This goes into the trailing curtain control mode and the screen N +
The contents of the frame memory up to line 1 + L + X are rewritten, and the rewriting of the frame memory is stopped for the subsequent lines and the previous image is held. Therefore, in the decoded screen N + 1, the contents of the screen N + 1 are present above the L + Xth line and the contents of the previous screen N remain below the L + X line. At the next N + 2nd screen time, the screen N + 2 is changed to the L + th screen.
The frame memory is rewritten up to 2X lines, and the rewriting of the frame memory is stopped for the subsequent lines. At the (N + 3) th screen time, the screen N + 3 is rewritten to the (L + 3X) th line in the frame memory and stopped. After all, in the screen time N + 1, N + 2, N + 3, respectively L + X, L
The frozen screen N remains below the + 2X, L + 3X lines, and the moving image that has undergone normal rewriting is supplied above the L + X, L + 2X, L + 3X lines, and the rewriting area expands over time, as if from the top to the bottom. It seems to have started the curtain. The speed at which the curtain is drawn is defined by X line / screen.

(Example) Next, an example of the present invention will be described with reference to FIG.

The moving image signal input from the input terminal 1000 is supplied to the scene change detection circuit 10, the synchronization signal generation circuit 12, and the delay circuit 22 via the line 1010.

Here, the scene change detection circuit 10 will be described with reference to FIG. The moving image signal input from the line 1010 is the moving vector detection circuit 31, the frame memory 32, and the delay circuit 38.
Is supplied to. The motion vector detection circuit 31 uses both the motion picture signal supplied from the line 1010 and the motion picture signal supplied from the frame memory 32 to calculate a motion vector in block units or pixel units. The motion vector thus generated is output to the variable delay circuit 33 and the variable delay circuit 18 shown in FIG. This motion vector detection method is described, for example, by Ninomiya in "Motion Compensation in Interframe Coding" (Journal of the Institute of Electronics and Communication Engineers, VOL.
J64-BNO.1 Sho 56 pp. 24-31, hereinafter referred to as "reference 2") can be used.

The frame memory 32 can store a moving image signal for about one screen, and outputs an image signal about one screen time before to the motion vector detection circuit 31 and the variable delay circuit 33. The variable delay circuit 33 variably delays the moving image signal supplied via the line 1213 according to the optimum prediction method supplied via the line 1018, and as a motion compensation prediction value, to the subtractor 34 via the line 1314. Output. The delay circuit 38 delays the input signal by the time required for the motion vector detection circuit 31 to calculate the motion vector and the variable delay circuit 33 to generate the motion compensation predicted value, and outputs the delayed signal to the subtractor 34. The subtractor 34 subtracts the motion compensation prediction value from the delayed input signal supplied via the line 1814 to obtain the motion compensation prediction error, and outputs it to the absolute value calculator 35.
The absolute value calculator 35 takes the absolute value of the motion compensation prediction error,
Output to adder circuit 36 via line 1516. The adder circuit 36 is
It resets with the head information of the frame supplied from the line 1210, starts adding the motion compensation prediction error supplied via the line 1516, and outputs the result to the comparator 37. When the absolute value of the motion compensation prediction error added from the beginning of the frame exceeds the predetermined threshold value, the comparator 37 outputs the code 1 to the control circuit 11 via the line 1011 as a scene change.

Return to FIG. The synchronizing signal generating circuit 12 separates and generates a frame pulse from the input moving image signal. For example, 1 at the beginning of each frame, 0 otherwise is controlled through the line 1210 to the control circuit 11, line counters 13 and 14, scene change. Output to the detection circuit 10.

The control circuit 11 sends a reset signal of the line counter at the beginning of the screen based on the frame pulse input from the line 1210 when the code 1 or the occurrence of a scene change is input via the line 1011. Line counter 13 through 1113
Output to. The line counter 13 is reset by a reset signal input via the line 1113, counts up by a value (X in the description of FIG. 1) determined by the frame pulse input via the line 1210, and the line 1315. Through the comparator
Output to 15.

The line counter 14 is a counter that is reset by a frame pulse supplied via the line 1210 and counts up by 1 in one line time, and outputs its value to the comparator 15 via the line 1415.

The comparator 15 compares the contents of the lines 1315 and 1415 and outputs the code 1 to the frame memory 25 via the line 1525 when the line 1415 is smaller and the code 0 otherwise. The delay circuit 22 delays the input signal by the time required to generate the code supplied from the comparator 15 and outputs the delayed signal to the frame memory 25. The frame memory 25 stores the image by rewriting the memory when the code 1 is input from the line 1525 and stores the code 0.
When is input, the rewriting of the memory is stopped and the previous image is retained. The output of the frame memory 25 is the subtractor 17
Is supplied to. The subtractor 17 subtracts the prediction signal supplied via the line 1817 from the input signal supplied from the frame memory 25 to calculate a prediction error signal, and outputs the prediction error signal to the quantizer. The quantizer 19 quantizes the prediction error signal, and the code converter
Output to 23 and adder 21. The adder 21 adds the output of the quantizer 19 and the prediction signal supplied via the line 1817 to generate a locally decoded signal, and outputs it to the frame memory 20. The frame memory 20 stores and delays the locally decoded signal for about one frame time, and outputs this to the variable delay circuit 18. The variable delay circuit 18 variably delays the output of the frame memory 20 according to the motion vector (optimal prediction method) supplied via the line 1018, and uses the line 1817 as a prediction signal.
To the subtractor 17 and the adder 21 via. Code converter 23
Converts the output of the quantizer 19 into a code suitable for transmission such as Huffman code and outputs the code to the buffer memory 24. The buffer memory 24 outputs the output of the code converter 23 to the transmission path 2000 at a constant transmission speed.

(Effects of the Invention) According to the present invention, as described above, in the frame memory of the input so as to smoothly switch annoying image quality deterioration such as encoding stop to the screen in the middle of the screen that often occurs when the contents of the screen greatly change. This can be avoided by rewriting, and a fine encoding mode can be applied to the encoded area, so that a decoded image with high image quality is provided. Further, in the present invention, the method of detecting a scene change is not limited, and the speed of curtaining is not particularly limited. It is also possible to start the curtain from a plurality of places. As described above, the present invention has great significance for practical use.

[Brief description of drawings]

FIG. 1 is a diagram showing the principle of curtain control. FIG. 2 is a diagram showing the principle of motion vector detection by the block matching method. FIG. 3 is a diagram showing the principle of scene change detection,
FIG. 4 is a diagram showing an embodiment of the present invention, and FIG. 5 is a diagram showing an embodiment of a scene change detection circuit. In the figure, 1000 is an input terminal, 10 is a scene change detection circuit, 11
Is a control circuit, 12 is a synchronizing signal generating circuit, 13 and 14 are line counters, 15 and 37 are comparators, 17 and 34 are subtractors, 18 and 33 are variable delay circuits, 19 is a quantizer, and 20 and 25, 32 is a frame memory, 22 and 38 are delay circuits, 23 is a code converter, 24 is a buffer memory, 31 is a motion vector detection circuit, 35 is an absolute value calculator, 36 is an addition circuit, 37 is a comparator, 2000 is Output end, respectively.

Claims (2)

[Claims] 1. A moving image signal input from an Nth screen to an Nth screen.
+1, N + 2 ... N + M, when encoding a moving image signal that changes with time, it is determined whether the N + 1th screen has changed significantly from the Nth screen, and if it is determined that the determination has changed significantly, A screen obtained by rewriting a certain area of the Nth screen with the N + 1th screen is set as a moving image signal to be coded, is coded by using the correlation with the Nth screen which is the screen of the previous time, and then the constant The region is gradually widened with the passage of time, and the widened region is rewritten on the N + 2, N + 3 ... N + M screens in time order to be a moving image signal to be encoded, which is to be encoded at the previous time. A method for encoding a moving image signal, which is characterized in that the encoding is performed by using the correlation with the moving image signal. 2. A judging means for judging whether or not the input moving image signal greatly changes the contents of the screen from the moving image signal of the previous time, and a synchronizing signal generating means for generating a synchronizing signal from the input moving image signal. A delay unit that delays the input moving image signal by a predetermined time; a first counting unit that counts the screen address in one frame time of the screen in synchronization with the synchronization signal; When it is determined that the output has significantly changed the contents of the screen, an address for designating a predetermined area for each screen is output in synchronization with the synchronization signal so that the predetermined area gradually expands. Second counting means, comparing means for comparing the addresses of the first counting means and the second counting means and outputting a control signal for controlling rewriting of the certain area; Out A storage unit that rewrites and stores a certain area of the screen of the previous time on the screen of the generated moving image signal and outputs this as the moving image signal of the encoding target, and the moving image signal of the encoding target of the previous time code. An encoding device for a moving image signal, comprising: an encoding means for encoding by using a correlation with a moving image signal to be encoded.