JPS6212287A - Adaptive predictive encoding device for moving image signal - Google Patents

Abstract

PURPOSE:To eliminate a decrease in the efficiency of an interframe predictive encoding system by selecting predictive encoding which uses a background memory for an uncovered background and predictive encoding which uses interframe correlation for other case. CONSTITUTION:The uncovered background is area-separated from an input signal by an uncovered background separating circuit 30 and a selector 32 selects a predicted background value. The input signal is delayed through a delay circuit 36 by a time required for the area separation and outputted to a subtracter 31, which subtracts the predicted value to generate a predicted error signal, which is outputted to an output terminal 300 and an adder 35. The adder 35 adds a predicted error signal to a predicted signal to generate a local decoded signal and an interframe predicting circuit 33 generates and outputs the predicted signal which uses the interframe correlation to the selector 32. A background predicting circuit 34 generates a predicted signal for a background including the uncovered background and outputs it to the selector 32.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention relates to digital compression transmission technology for television signals. The present invention relates to an adaptive predictive coding technique for video signals that maintains efficiency.

(Conventional Technique t#) Conventionally, in order to highly efficiently compress and transmit television signals, etc., an interframe coding method using motion compensation has been used. This is described by Ninomiya et al. [Motion Compensated Interframe Coding Method] (Transactions of the Institute of Electronics and Communication Engineers (B), J63-B
, 11. pp, 1140-1147 (Sho 5l-11))
As announced in 2007, this is a technology that efficiently compresses bandwidth by utilizing the fact that video signals have a high interframe correlation.

(Problems to be Solved by the Invention) A moving image signal includes, in addition to a static region such as a background and a moving region created by a moving object, an uncovered background that newly appears from behind the moving object. In this region, there is no interframe correlation at all, so a predictive coding method that uses interframe correlation, such as the above-mentioned motion compensation coding method, has the drawback of significantly lowering efficiency.

An object of the present invention is to eliminate the decrease in efficiency of an interframe predictive coding method in an uncovered background.

(Means for Solving the Problems) According to the present invention, there is provided a means for performing motion compensation prediction on a moving image signal such as a television signal and calculating the prediction error, and a first evaluation for evaluating the motion compensation prediction error. means, the same (means for calculating a prediction error by performing interframe prediction on a video signal, a second evaluation means for evaluating the interframe prediction error, and a means for calculating the first evaluation result and the second evaluation result) means for storing, third evaluation means for evaluating the stored first and second evaluation results, means for performing predictive encoding using inter-frame correlation, means for performing background predictive encoding using background memory; According to the third evaluation result, an adaptive predictive coding device for a moving picture signal is obtained, which is characterized by having means for selecting predictive coding using a background memory for uncovered backgrounds.

(Operation) Next, the principle of operation of the present invention will be explained. In the following description, a television signal will be used to represent a moving image signal.

First, a method for accurately region-separating the uncovered background, which is important in the present invention, will be explained.

First, a motion compensation prediction error value and a frame difference value (interframe prediction error) are determined. As the motion-compensated prediction error, for example, the prediction error calculated by the motion-compensated predictive coding method described in the document by Ninomiya et al. mentioned above can be used as is. That is,
Divide the screen into blocks of a certain size, find the block with the highest correlation from the previous frame signal stored in the frame memory, and calculate the difference value of pixels at the same position between corresponding blocks as the motion compensation prediction error. value.

The method of obtaining the motion compensated prediction error value is not limited to the method described above, and may be obtained using other methods. The frame difference value is obtained by calculating the difference between the current pixel and a pixel at the same position in the frame memory.

In the following explanation, it is assumed that the motion compensation prediction error value and the frame difference value have already been determined.

In FIG. 2, ■, ■, ■ are four consecutive frames Fl, F2. F3. The position in each frame when an oblique oval moving object moves in parallel to the right at F4 and stops at F3 is overwritten on a single diagram.

The dashed line, dashed-dotted line, and solid line are Fl, F2°F3, respectively.
It represents the position of the object at ■, ■, ■. The animal body is
Frame Fl, F2. It moved parallel to F3 and came to the positions of ■, ■, and ■, respectively, and stopped at F3, so F
In 4, it is in the same position as in F3. Next, in Figure 2, [
The parts indicated by 11 to [6] are the moving areas in each frame,
Classify whether it is a stationary area. [Moon is Fl,
F2. F3. The background through F4 is the static area, [2] is Fl, the uncovered background between F2, the background after F2, [3] is Fl, F2
In between, there is a dynamic region, F2. Between F3 is the uncovered background, after F3 is the background, [4] (the part that overlaps when the ellipses are in the same position as ■ and ■) is Pi, F2
．． F3. Dynamic area through F4, [5] (same as ■
The part that overlaps when and ■ are in the same position) is Fl, F2
F2. F3 is the dynamic area, after F3 is the stationary area, [6
] (the part excluding the previous [5] at position ■) is Fl,
Between F2, background, F2. The area between F3 and F3 is covered with moving objects, and the area after F3 is a stationary area.

The motion compensation prediction error values and frame difference values obtained for pixels in the vicinity of [11 to [6] in Figure 2 are compared in magnitude with an appropriate threshold T, and the error values and difference values between F1 to F4 are calculated. If we examine the changes qualitatively, we can see the results shown in Figure 3. FO in Figure 3
is the screen before Fl.

In the case of the movement shown in Figure 2, as shown in Figure 3, the uncovered background ([2] and [3]
), the motion compensation prediction error changes from small to large, and then the frame difference value changes from large to small. If the object is moving continuously, even if it is only the time transition of the frame difference value, other regions ([11, [
4], [5], [6]), but if the object stops, the frame difference value will be similar to that of the uncovered background even in the regions [4], [5], [6]. , indicating a transition from large to small (indistinguishable). Even in such cases, it is sufficient to examine the time transition of the motion compensated prediction error. When the object is stationary, [4], [5]
, [6] are continuously small, and the motion compensated predicted value in the actual uncovered background becomes large once before the frame difference value transitions from large to small as described above. They can be distinguished because they exhibit different behavior.

Next, the operation of the present invention in the uncovered background will be explained using FIG.

FIG. 1 is a diagram showing luminance amplitude values and frame difference values for the same object movement as in FIG. 2. (A) shows the luminance amplitude value at F2 of the cross section indicated by the broken line in Figure 2, with the horizontal axis being the position (X
(B) is the brightness amplitude value at F3. (C) is the frame difference value at F3, that is, (B) - (A). [2] corresponds to the area [2] in FIG. 2 mentioned above, and is the uncovered background newly appearing in (B), that is, F3.

As can be seen from this figure, the uncovered background is an area where there is no correlation between frames, so by taking the interframe difference value, the image signal of the current frame (
(B) [2], uncovered background) and the image signal from one frame before ((A) [2], moving area)
This is not efficient as it includes both information. Therefore, for the uncovered background, by performing background prediction using the background memory, it is possible to improve the encoding efficiency in this area. As for the structure of the background memory and the prediction method using it, for example, "Interframe coding method using motion compensation/background prediction" by Kuroda et al.
(The Journal of the Institute of Electronics and Communication Engineers, January 1985, Volume J68-B, No. 1, Pages 77 to 84) can be used.

(Example) Next, the present invention will be explained in detail.

FIG. 4 is a diagram illustrating the uncovered background separation circuit.

A moving image signal is input to the input terminal 100, and the delay circuit 1
3 and 16, frame memories 10 and 11, and a motion vector detector 12. The delay circuits 13 and 16 operate only for the time necessary to calculate the interframe predicted value via the frame memory 10 and the delay circuit 14, and the motion compensated predicted value via the frame memory 10 and the variable delay circuit 15. The input signals are delayed and provided to each subtractor 517.18 via lines 1317.1618. The frame memory 10 is capable of storing approximately one frame of moving image signals. The delay circuit 14, together with the frame memory 10, has enough delay to delay the input signal by one frame, and the input signal delayed by one frame is output to the subtracter 17 via a line 1417. . The subtractor 17 is the line 1317
1 input from line 1417 and 1 input from line 1417.
A frame difference value is calculated using the frame delay signal and outputted to the comparator 19. The comparator 19 compares the frame difference value and the threshold T, and outputs a code 1 to the memory 21 if the absolute value of the frame difference value is larger, and a code 0 if it is smaller.

The frame memory 11 is a memory for detecting motion vectors, and stores input signals for one frame. For the motion vector detection 512, the motion vector detection circuit described in the above-mentioned document by Ninomiya et al. can be used. That is, a motion vector (optimal prediction method) is calculated using the block matching method using an input signal supplied from the line 100 and a signal from approximately one frame before supplied via the line 1112, and this is calculated by the line 1112. 12
15 to the variable delay circuit 15. Variable delay circuit 15 delays the frame memory output supplied via line 1015 according to the optimal prediction scheme and outputs it to subtracter 18 as a motion compensated predicted value. The subtractor 18 is connected to the line 16
The motion compensated prediction error value is calculated by subtracting the input signal provided via line 18 and the motion compensated prediction value provided via line 1518 and output to comparator 20 . Comparator 20
compares the motion compensation prediction error and the threshold T,
For example, if the absolute value of the motion compensation prediction error value is larger, a code 1 is output to the memory 22, and if it is smaller, a code 0 is output to the memory 22.

The memories 21 and 22 are capable of storing several frames of the comparison results between the frame difference value and the threshold value, and the comparison result between the motion compensation prediction error value and the threshold value, respectively.

The determination circuit 23 reads out the comparison result between the frame difference value stored in the memory 21 and the threshold value, and the comparison result between the motion compensation prediction error value stored in the memory 22 and the threshold value, and reads out the comparison result between the frame difference value stored in the memory 21 and the threshold value, and reads out the result of the comparison result between the motion compensation prediction error value stored in the memory 22 and the threshold value, and reads out the result of the comparison between the frame difference value stored in the memory 21 and the threshold value, and reads out the comparison result between the motion compensation prediction error value stored in the memory 22 and the threshold value, and reads out the comparison result between the frame difference value stored in the memory 21 and the threshold value, and reads out the comparison result of the magnitude difference between the frame difference value stored in the memory 21 and the threshold value, and reads out the comparison result of the magnitude difference between the motion compensation prediction error value stored in the memory 22 and the threshold value. , it is determined whether the corresponding pixel (or block) is in the uncovered background, and if it is in the uncovered background, a code 1 is output to the output terminal 200, and if not, a code O is output to the output terminal 200. This can be easily realized using a read-only memory (ROM) that uses each comparison result as an address input.

FIG. 5 shows an example of the determination logic. In the case of Fig. 5, the comparison result of the current motion compensation prediction error is input to address A3.
If the comparison result of the motion compensation prediction error before 1 screen time for 2, 2 screens for A1, and 3 screen times for AO is greater than the threshold, set it to 1.
If it is small, it is expressed as 0 and input. Similarly, the comparison results of frame difference values are inputted into A7 to A4 from the current time three screen times ago. Uncovered background is a case where the motion compensation prediction error changes from small to large and is currently small one frame before the frame difference value changes from large to small.
When the address is 116 in decimal notation, the output is 1 because it is an uncovered background.

FIG. 6 shows an embodiment of an interframe adaptive predictive coding device using the uncovered background separation circuit shown in FIG. 4.

An input moving image signal input from the input terminal 100 is supplied to an uncovered background separation circuit 30 and a delay circuit 36. Uncovered background separation circuit 30
separates the uncovered background from the input signal, and outputs code 1 to the selector 32 if the background is uncovered, and outputs code O if not. Selector 32 selects the background predicted value supplied via line 3432 when code 1 is input via line 200, and selects the background predicted value supplied via line 3432 when code 0 is input. 36 supplied. The uncovered background separation circuit 30 separates the uncovered background from the input signal, and gives a code of 1 when the background is uncovered, and a code of 0 otherwise.
is output to the selector 32. The selector 32 is connected to the line 200
When code 1 is input via line 3432, select the background predicted value provided via line 3432, and when code O is provided, use the interframe correlation provided via line 3332. Select the predicted value and send it to adder 3 via line 3235.
5 and output to the subtracter 31. The delay circuit 36 delays the input signal by the time necessary for the uncovered background to be separated into regions in the uncovered background separation circuit 30 and outputs this to the subtracter 31 . The subtracter 31 extracts a line 32 from the delayed input video signal.
The predicted value provided via 35 is subtracted to produce a prediction error signal which is output via line 3130 to output 300 and adder 35. Adder 35 adds the prediction error signal and the prediction signal supplied via line 3235 to create a locally decoded signal, which is sent to interframe prediction circuit 33 via line 3530.
is output to the background prediction circuit 34. Interframe prediction circuit 3
3 generates a prediction signal using inter-frame correlation; for example, the motion-compensated inter-frame prediction circuit described in Ninomiya's literature mentioned above can be used, and the motion-compensated inter-frame prediction signal is generated by line 3332. The output signal is output to the selector 32 via the input signal. The background prediction circuit 34 generates a prediction signal for the background including the uncovered background.
For example, the background prediction circuit described in the aforementioned Kuroyama et al. paper can be used and outputs a background prediction signal to the selector 32 via line 3432.

(Effects of the Invention) As described above, according to the present invention, an inter-frame adaptive predictive coding device with higher efficiency for the uncovered background that newly appears from behind a moving object is realized, and the significance of putting the present invention into practical use is It's large.

[Brief explanation of drawings]

Figure 1 shows the change in luminance amplitude and frame difference values when an object moves parallel to the screen, and Figure 2 shows the position when the object moves parallel to the screen and then stops. Figure 3 is a diagram showing the results of a comparison between the motion compensation prediction error value of each region in Figure 2 and the threshold value of the frame difference value, and Figure 4 is a diagram showing the implementation of the uncovered background separation circuit. FIG. 5 is a diagram showing an example; FIG. 5 is a diagram showing the logic of the determination circuit shown in FIG. 4; and FIG. 6 is a diagram showing an embodiment of an interframe adaptive coding apparatus using the present invention. In 1m, 10.11 frame memory, 12 motion vector detector, 13.14.16.36 delay circuit, 15 variable delay circuit, 17.18.31 subtractor, 19.20 comparator, 21 .22 is a memory, 23 is a determination circuit, 30 is an uncovered background separation circuit, 32 is a selector, 33 is an interframe prediction circuit, 34 is a background prediction circuit,
35 represents an adder, 100 an input end, and 300 an output end. O 1 Figure O 2 Figure

Claims (1)

[Claims] Means for performing motion compensation prediction on a moving image signal such as a television signal and calculating a prediction error thereof; first evaluation means for evaluating the motion compensation prediction error; means for calculating an error; second evaluation means for evaluating the inter-frame prediction error; means for storing the first evaluation result and the second evaluation result; and the stored first and second evaluation results. A third evaluation means for detecting a newly appearing background (uncovered background) after the movement of a moving object by comparing the images, means for predictive coding using inter-frame correlation, and background predictive coding using background memory. According to the third evaluation result, predictive coding using background memory is selected for the uncovered background, and predictive coding using interframe correlation is selected for other cases. An adaptive predictive encoding device for a moving picture signal, characterized in that it has means.