JPS61201568A - Adaptive coding system - Google Patents

Abstract

PURPOSE:To apply coding at scene change without decreasing the coding efficiency by selecting a forecast signal having a larger correlation in a correlation discriminating section, adding the result to a coding section, obtaining a difference with an input signal and applying coding to the difference. CONSTITUTION:The 1st forecast signal close to an input signal is formed by a vector quantizing section 2, further a correlation discriminating section 4 from which the 2nd forecast signal where the direction and magnitude of a pattern are compensated by a movement compensation forecast section 3 is formed discriminates the quantity of correlation among the input signal and the 1st and 2nd forecast signals and then the presence of scene change is discriminated. Then in the coding section 1, a forecast signal discriminated as large correlation is used and the difference with the input signal is obtained and the difference is coded.

Description

[Detailed Description of the Invention] [Industrial Field of Application] The present invention provides a method for band compression transmission of image signals so that encoding efficiency does not decrease even when inter-frame correlation is small due to scene changes, etc. The present invention relates to an adaptive coding method.

[Conventional technology]

In band compression transmission methods for image signals, there are two methods: intra-frame encoding, which encodes intra-frame differences such as differences between pixels on scanning lines, such as between adjacent pixels, and differences between scanning lines; An interframe encoding method that encodes an interframe difference between pixels with respect to a frame, and an interframe composite encoding method that combines these are known.

The intra-frame encoding method encodes the difference between pixels within the same frame, so for monotonous screens, the correlation between pixels increases and the encoding efficiency increases.
In the case of a screen that includes many changing parts such as border lines, the correlation between pixels is small, so the encoding efficiency is not very high.

Furthermore, in the interframe coding method, when the current frame and the previous frame are almost the same, such as in a still screen, the interframe difference is zero or a value close to zero, so the coding efficiency is high. Furthermore, since the interframe composite encoding method calculates and encodes interframe differences and intraframe differences, high encoding efficiency can be obtained even for normal moving screens.

Also known is a motion compensation method that detects the direction and magnitude of screen motion when determining the interframe difference, and compensates the interframe prediction signal in accordance with the direction and magnitude.

[Problem that the invention seeks to solve]

When switching image signals from multiple TV cameras,
When a scene changes, such as when the viewing range or direction of a TV camera changes suddenly, the correlation between the frames before and after the change is close to zero, and the coding efficiency of the interframe coding method is significantly reduced. There is a drawback that it decreases.

Intra-frame encoding is not affected by scene changes, but the clearer the screen, the smaller the correlation between pixels.
The disadvantage is that it is difficult to increase the encoding efficiency. Furthermore, since the interframe composite coding method includes interframe coding, it has the disadvantage that the coding efficiency decreases at scene changes.

The present invention aims to improve the above-mentioned conventional drawbacks, determine scene changes, etc. based on correlation between frames, and encode image signals using an optimal encoding method.

[Failure to solve the problem]

The adaptive encoding method of the present invention includes an encoding section 1, a vector quantization section 2, a motion compensation prediction section 3, and a correlation determination section 4, and a first prediction signal from the vector quantization section 2 with respect to an input signal. , a correlation determination unit 4 determines the correlation between the second prediction signal from the motion compensation prediction unit 3 and the human signal;
The predicted signal with the larger correlation is selected and added to the encoding unit 1, the difference with the input signal is determined, and the difference is encoded.

[Effect]

The vector quantization unit 2 forms a first prediction signal close to the input signal, the motion compensation prediction unit 3 forms a second prediction signal compensated for the direction and magnitude of screen motion, and the correlation determination unit 4 In the input signal and the first and second
The magnitude of the correlation with the predicted signal is determined, and the presence or absence of a scene change or the like is determined accordingly. Then, in the encoding unit 1, a difference from the input signal is determined using the predicted signal determined to have a large correlation, and the difference is encoded.

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention, in which a digitized image signal is transmitted from an input terminal 5 to an encoding unit 1, a vector quantization unit 2, a motion compensation prediction unit 3, and a correlation determination unit 4.
added to. The encoding unit 1 has a configuration that performs intraframe encoding and interframe encoding, and the encoded signal is transferred from an output terminal 6 to a transmitting unit (not shown) including a buffer memory and the like. The vector quantization section 2 forms a first prediction signal close to the input signal by encoding and decoding the human input signal, and corresponds to a feedforward prediction signal generation section.

The motion compensation prediction unit 3 includes a frame memory storing an interframe prediction signal, a motion vector detection circuit, etc., and calculates screen motion by comparing the input signal with a block of a predetermined size one frame before. Detect the direction and size,
forming a second predicted signal close to the input signal;
This corresponds to a feedback type prediction signal generation section.

The correlation value between the first predicted signal from the vector quantization unit 2 and the input signal corresponds to the intra-frame correlation value, and the correlation value between the second predicted signal and the input signal from the motion compensation prediction unit 3 corresponds to the intra-frame correlation value. This corresponds to the intercorrelation value, and the correlation determination unit 4
It has a configuration in which the magnitude of these correlation values is determined in units of blocks of m lines x n pixels, and control is performed to select the first or second prediction signal and add it to the encoding unit l. For example, in a scene change, the interframe correlation becomes small, so the correlation value between the input signal and the second predicted signal is
is smaller than the correlation value between the input signal and the first predicted signal,
The correlation determination unit 4 determines the magnitude of such a correlation value, and
The first prediction signal from the vector quantization unit 2 is controlled to be added to the encoding unit 1. As a result, the encoding unit 1 calculates the difference between the input signal and the first predicted signal and encodes the difference, so that encoding equivalent to intra-frame encoding is performed. Even when the inter-frame correlation is small, it is possible to prevent the encoding efficiency from decreasing.

Due to a scene change, encoding is performed using the first predicted signal as described above, and as the transition from a transient state to a steady state occurs, the interframe correlation increases, and the input signal and first predicted signal are The correlation value between the input signal and the second
Since the correlation value with the predicted signal becomes large, the correlation determination unit 4
controls so that the second prediction signal from the motion compensation prediction unit 3 is added to the encoding unit 1, so the encoding unit l calculates the difference between the input signal and the second prediction signal, and calculates the difference between the input signal and the second prediction signal. Since the data is encoded, encoding equivalent to interframe encoding is performed.

FIG. 2 is a detailed block diagram of an embodiment of the present invention, in which the same reference numerals as in FIG. 1 indicate the same parts, 10 is a delay circuit, 1
1 is a subtracter, 12 is a quantization unit that quantizes the output signal of the subtracter 11 in predetermined steps, 13 is an adder, 14 is a selector, 15 is an encoder, 16 is a decoder, 17 is a frame memory, 18 1 is a motion vector detection circuit, 19 is a variable delay circuit, 20 is a selector, 21 is an output terminal for a quantized output signal,
Reference numeral 22 represents an output terminal for a determination signal, and 23 represents an output terminal of the selector 20. Means for encoding a difference and means for determining an intra-frame difference are not shown for the sake of simplicity.

A digitized image signal is applied to the input terminal 5 as an input signal a, and is applied to the vector quantization section 2, the motion compensation prediction section 3, and the correlation determination section 4. The vector quantization unit 2 includes an encoder 15 and a decoder 16, encodes the input signal a according to a predetermined algorithm, decodes it, and outputs a first predicted signal close to the input signal a.

The motion compensation prediction unit 3 also includes a frame memory 17 and a motion vector detection circuit 18 . Consisting of a variable delay circuit 19, etc.,
The output signal of the adder 13 is added to the frame memory 17 as an interframe prediction signal for the input signal a of the next frame. The motion vector detection circuit 18 also examines the correlation between the input signal a and the block correspondence of a predetermined size in the frame memory 17, and obtains a motion vector signal indicating the direction in which the correlation is greatest and its magnitude. The variable delay circuit 19 is controlled in response to this motion vector signal, and outputs a second prediction signal C that compensates for the direction and magnitude of screen motion.

The correlation determination unit 4 receives the input signal a, the first predicted signal C, and the second predicted signal C, and calculates the correlation value between the input signal a and the first predicted signal C, and the input signal a and the second predicted signal C. The magnitude of the correlation value between the predicted signal C and the correlation value is determined, and the determined signal e is applied to the selector 14.20 and the output terminal 22. The selector 14 selects the first prediction signal or the second prediction signal according to the determination signal e.
The predicted signal C is selectively outputted and added to the subtracter 11, and the difference between it and the input signal a compensated for the delay time in the vector quantizer 2 etc. is determined by the delay circuit 10, and the output signal of the subtracter 11 is The quantization unit 12 also performs quantization, and the quantized output signal d is applied from an output terminal 21 to an encoder (not shown). The quantized output signal d is added to the output signal of the selector 14 by the adder 13 and added to the frame memory 17 as an interframe prediction signal for the input signal a of the next frame.

Further, the selector 20 selects the vector quantization signal f from the vector quantization section 2 or the motion vector signal g from the motion compensation prediction section 3 according to the determination signal e.

The image signal is selected and sent to the receiving side together with the determination signal e, which becomes information indicating the encoding format of the encoded image signal.

At the scene change, as described above, the correlation determination unit 4 determines that the correlation value between the input signal a and the first predicted signal is greater than the correlation value between the input signal a and the second predicted signal. Based on the determination signal e, the selector 14 selects and outputs the first prediction signal S and applies it to the subtracter 11. Also, the selector 20 is the vector quantizer 2
The vector quantized signal f is selected and outputted to the output terminal 23. The selector 14 selects and outputs the first predicted signal, the subtracter 11 calculates the difference from the input signal a, and the difference is quantized and then encoded, which corresponds to intra-frame encoding. It is possible to prevent deterioration in encoding efficiency in scene changes where encoding is performed and the correlation between frames is very small.

In addition, in the case of a screen with normal movement, the correlation value between input signal a and second predicted signal C is larger than the correlation value between input signal a and first predicted signal S, so correlation determination is difficult. The selector 14 selects the second prediction signal C based on the determination signal e from the section 4.
is selectively output and added to the subtracter 11, and encoding equivalent to interframe encoding is performed.

〔Effect of the invention〕

As explained above, the present invention includes the encoding section 1, the vector quantization section 2, the motion compensation prediction section 3, and the correlation determination section 4, and the first prediction from the vector quantization section 2 for the input signal a. When the signal is received, a correlation determination unit 4 determines the correlation with the second prediction signal C from the motion compensation prediction unit 3,
The predicted signal with the larger correlation is selected and added to the encoding unit 1, and the difference between this predicted signal and the input signal a is encoded, and scene changes, etc. are easily determined by the correlation determining unit 4. When it is determined that there is a scene change, the second prediction signal C formed based on the input signal a of the current frame is not used, and the second prediction signal C formed based on the input signal a of the current frame is used. 1 prediction signal is used to find the difference, and the coding equivalent to intra-frame coding does not reduce coding efficiency (it has the advantage of being able to perform coding at scene changes). .

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of the invention, and FIG. 2 is a detailed block diagram of an embodiment of the invention. ■ is an encoding unit, 2 is a vector quantization unit, 3 is a motion compensation prediction unit, 4 is a correlation determination unit, 5 is an input terminal, 6 is an output terminal,
10 is a delay circuit, 11 is a subtracter, 12 is a quantization unit, 13
is an adder, 14.20 is a selector, 15 is an encoder, 16
1 is a decoder, 17 is a frame memory, 18 is a motion vector detection circuit, and 19 is a variable delay circuit. A block diagram of an embodiment of the present invention Fig. 1 9 A detailed diagram of the present invention Fig. 2

Claims (1)

[Claims] An encoding section, a vector quantization section, a motion compensation prediction section, and a correlation determination section are provided, and the correlation between the input signal and the first predicted signal from the vector quantization section and the correlation between the input signal and the motion compensation prediction section are provided. The correlation with the second predicted signal is determined in the correlation determining section, and the predicted signal with the larger correlation is selected and added to the encoding section, and the encoding section determines the correlation with the input signal. An adaptive encoding method characterized by calculating a difference and encoding the difference.