JPS62209984A - Method and apparatus for encoding picture signal - Google Patents

Abstract

PURPOSE:To prevent the encoding from being stopped by expanding a dead zone or a step of a quantizer just after scene change and reducing the step gradually with time thereby suppressing the occurrence of excess information due to the difference of the amplitude being not zero at the scene change. CONSTITUTION:In applying forecast encoding or conversion encoding to an moving picture signal, the change in the picture content between patterns is evaluated and when it is discriminated that the content of picture between the patterns is changed remarkably, at least one of the dead zone and the step of the quantizing characteristic with respect to the forecast error or conversion coefficient is expanded, and the reduced gradually with time and it is restored to a predetermined dead zone or step. That is, a scene change detector gives a scene change signal to an inter-frame coder when a scene change takes place. The inter-frame coder expands the dead zone of the quantizer when the scene change signal is given, reduces it gradually with time and restores the processing into the normal quantization.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image signal encoding method and apparatus.

(Shorai's technology) Normally, in inter-frame encoding using inter-screen correlation, the second
As shown in the figure, the output of the interframe encoder is variable-length encoded by a variable-length encoder, and the generation speed of the encoded information contained in the variable-length encoder is matched with the transmission path speed. It is temporarily stored in the buffer memo IJ-. However,
When the screen content changes significantly, the correlation between screens decreases significantly, so encoding using this will generate a large amount of information. In conventional interframe encoders, when a large amount of information is generated and the buffer memory is about to overflow, a stop signal is usually given to the interframe encoder to stop encoding.

(Problem to be solved by the invention) In conventional interframe encoding using interframe correlation, excessive information is generated when the screen content changes significantly (hereinafter referred to as a scene change). If this occurs and the buffer memory is about to overflow,
Even if the encoding was stopped in the middle of the screen, the previous screen and the new screen that was being encoded would appear to overlap unnaturally in one screen, or the movement of the moving image would be discontinuous, resulting in a picture-story show. This was very distracting as it caused a lot of movement.

(Means for Solving the Problems) According to the first invention of the present application, when performing predictive encoding or transform encoding of a moving image signal, changes in image content between screens are evaluated, and changes in image content between screens are evaluated. When it is determined that the image content has changed significantly, at least one of the dead zone or step of the quantization characteristic for the prediction error or the transform coefficient is expanded and gradually reduced over time to a predetermined value. A method for encoding an image signal is obtained, which is characterized in that the original dead sea 7 or step is returned to the original dead sea 7 or step.

According to a second invention of the present application, there is also provided a device for predictively encoding an input moving image signal, comprising means for evaluating whether or not a significant change has occurred in the content of an image between screens; means for generating a difference signal between the moving image signal and the pre-6111 signal; means for quantizing the difference signal; and when performing the quantization, the output of the evaluation means indicates that there has been a significant change in screen content; When the dead zone or step of the quantization characteristic is temporarily expanded, the dead zone or step of the characteristic is gradually reduced over time to restore the predetermined original quantization characteristic. quantization control means, means for generating a locally decoded signal using the output of the quantization means and the prediction signal, means for generating the prediction signal using the locally decoded signal, converting the output of the quantization means into a variable length code. There is obtained an image signal encoding device characterized by comprising: means for encoding an image signal.

Furthermore, according to the third invention of the present application, there is provided an encoding device that encodes an input moving image signal by transform encoding, and the encoding device determines whether or not a significant change has occurred in the content of the image between screens. means for evaluating, means for generating a difference signal between the input moving image signal and the predicted signal, means for performing orthogonal transformation on the difference signal, means for converting the output of the orthogonal transformation means,
When performing nuclear identification, if the output of the evaluation means indicates that there has been a significant change in the screen content,
quantization control means for once enlarging at least one of the dead zone or step of the participantization characteristic and gradually reducing the dead zone or step over time to return to a predetermined original quantization characteristic; means for orthogonally inversely transforming the output of the quantizing means; means for generating a locally decoded signal using the output of the orthogonally inversely transforming means and the pre-1tllll signal; and generating the predicted signal using the locally decoded signal. An image signal encoding device is obtained, comprising: means for variable length encoding the output of the quantization means.

(Function) In the encoding control method of the present invention, as shown in FIG. 1, first, a see/change detector detects a see/change in which the image content changes significantly between screens from an input moving image signal. . For information on how to detect scene changes, see
Although not the subject of the present invention, an example of the paper by Miyahara is ``A farce and study of frame difference signal characteristics for band compression''.
(Eligibility theory (A), vOL56-A, N08.PP, 45
6-463 (August 1973) can be used. This method detects a significant change in the content of an image by determining the number of significant frame differences and the inter-frame correlation of this number. The scene change detector is
A see/change signal is provided to the interframe encoder when a change occurs. When the interframe encoder receives a 7-7 chip signal, it expands the dead zone of the quantizer, gradually shrinks it over time, and returns to normal quantization.

In the method of increasing the dead zone, for example, when a scene change occurs in a certain frame as shown in Fig. 3, the dead zone 7 of the quantizer is set to the width A in Fig. 4 for a period of N frames from that frame. During the next M frames, the dead zone of the quantizer is made slightly smaller than during the N frame time, as shown in B in FIG. After a further time, in the time domain of P, the dead zone of the quantizer is further reduced, and the width of the dead zone returns to the normal quantization characteristic of C.

During a scene change, a large number of differential signals with non-zero amplitudes were generated, causing the buffer memory to overflow, causing encoding to stop and deteriorating image quality. Therefore, when a sea/change occurs,
By expanding the dead zone of the quantizer and first suppressing excessive information generated by small-field difference signals, it is possible to prevent overflow of the buffer memory and thereby avoid stopping the encoding and maintain smooth motion. Also the fifth
As shown in the figure, it is also possible to perform coarse quantization by increasing the quantization step. For example, when a scene change occurs in a certain frame as shown in FIG. 38, the quantization step is coarsened for N frames from that frame, as shown in FIG. , when the time is N frames, the quantization step is made slightly finer and the characteristics shown in FIG. 5(B) are used. After a further time, in the time domain P, the quantization step is made finer and the normal quantization characteristic as shown in FIG. 5(C) is restored. By performing coarse quantization at the time of see/change in this way, the number of quantization steps can be reduced compared to when performing normal quantization, and short codes can be used even for large amplitude difference signals: , L-h'- allows you to change clothes to Akira 10ft 4m tiJ4's Masutsu winter north'0, -shi. Furthermore, it is also possible to combine the method of increasing the dead zone and the method of coarsening the quantization step.

(Example) Next, the present invention will be described in more detail with reference to Examples. 6th
The figure is a block diagram showing an embodiment of the second invention of the present application. An input moving image signal is supplied to a scene change detector 1 and a subtractor 2 via a line 101. The scene change detector l transmits the scene change signal to the quantizer 3 via a line 102 when the input moving image signal changes dramatically between screens and a scene change is detected.
supply to. The subtracter 2 subtracts the input moving image signal and the prediction signal supplied from the frame memory 5,
The difference signal is supplied as output to the quantizer 3 via line 201. The quantizer 3 converts the dead zone into a dead zone, for example, when a change occurs to perform quantization on the difference signal supplied from the subtracter 2 and a scene change signal is provided from the scene change detection D1. Figure 4 (A
), and after a while, shrink the dead zone to make it look like Figure 4 (B). Then, after a further period of time, Dead Sea 7 is returned to its normal value. Details of the quantizer 3 will be described later. The output of the quantizer 3 is supplied to an adder 4 and a variable length encoder 6. Adder 4 uses the difference signal supplied from quantizer 3 and the prediction signal supplied from frame memory 5 to reproduce the locally decoded signal. The locally decoded signal, which is the output of adder 4, is supplied to frame memory 5. Frame memory 5 delays the locally decoded signal by one frame time and supplies the predicted signal to subtracter 2 and adder 4 as output.

The variable length encoder 6 performs variable length encoding on the difference signal supplied from the quantizer 3 using an efficient code such as a Huffman code, matches the encoding speed and the transmission path speed, and outputs a line 601. The variable length code is supplied to the transmission path through the transmission line.

Next, referring to FIG. 7, a decoder for reproducing the original moving image signal from the output variable length code of the embodiment of FIG. 6 will be explained. A variable length code is supplied from the transmission path to the variable length decoder 7 via a line 701. The variable length decoder 7 performs variable length decoding while matching the speed of the transmission path and the speed of decoding. The difference signal reproduced by variable length decoding of the output of variable length decoder 7 is supplied to adder 8 . Eight adders perform decoding using the difference signal supplied from the variable length decoder 7 and the prediction signal supplied from the frame memory 9 to reproduce the original moving picture signal. The decoded signal output from adder 8 is supplied to frame memory 9 . It is also provided as the output of the decoder via line 801. Frame memory 9 is
The decoded signal supplied from the adder 8 is delayed by one frame time, and the predicted signal is supplied to the adder 8 as an output.

Next, the quantizer 3 will be explained in detail with reference to FIG. The quantizer 3 stores the difference signal supplied from the subtracter 2 via the line 201 in a read-only memory (Rsad 0
nly Memory: ROM) ROM-P
2O3, ROM-N302, ROM-M3Q3. The ROM-P3Ql stores in advance the basic 4: Childization characteristic. ROM-N302 is the fourth
A quantization characteristic with an enlarged dead zone as shown in A in the figure is stored in advance. The ROM-N 303 stores in advance a quantization characteristic having a dead zone as shown in B in FIG. Note that these ROMs use line 201 for address input.
The value of the quantized output is stored in correspondence with the address in order to provide the difference Σ signal supplied by the address. The output of each ROM, that is, the quantized output, is then supplied to a switch 306. Timer 305 supplies a timing signal indicating a frame to control circuit 304. Normally, the control circuit 304 applies a control signal to the switching unit 306 to select ROM-P2O1 and performs normal quantization.

When the change signal is given, the timer 305
Immediately after a scene change occurs, a control signal is sent to select the quantized ROM-N302 with an enlarged output dead zone based on the timing signal supplied from ROM-
A control signal for selecting M303 is given to the switch 306, and after a further time, a control signal is given to the switch 306 for selecting ROM-P3O1, which performs normal quantization. Switcher 30
6 selects a quantization ROM according to a control signal supplied from the control circuit 304, and outputs the quantizer 3.

Further, by storing coarse characteristics of quantization steps as shown in FIG. 5 in the ROM-N 302 and ROM-N 303, coarse quantization encoding can be performed. In this case, the ROM-N302 has the characteristics shown in FIG. 5(A),
The characteristics shown in FIG. 5 0) are stored in the ROM-N 303. The ROM is switched in the same way as in the previous embodiment, and when a scene change occurs, the coarse quantization ROM is switched.
-N302 is selected, and after a while ROM-, M
3Q3 is selected, and after some time, ROM-P2O
1 is selected. In this way, during the scene checker 72, by coarsening the quantization step, reducing the number of steps, and encoding with a short code, it is possible to prevent the generation of excessive information for large amplitude differences. Furthermore, it is also possible to use a combination of a method of widening the dead zone and a method of coarsening the quantization step. In this case, R
Store in the OM-N302 the dead zone with the width A in Figure 4 and the step with the characteristics shown in Figure 5 (A).
M3Q3 can be handled by storing the width of the dead zone as shown in B in FIG. 4 and the characteristic of the step as shown in FIG. 5 (b).

Next, an embodiment of the third invention of the present application will be described with reference to FIG. This embodiment is an encoding device that combines orthogonal transformation and predictive encoding. An input moving image signal is sent to a sea/change detector 1 and a subtractor 2 via a line 101.
supplied to The scene change detector 1 converts the scene change signal into a line 1 when the input moving image signal changes significantly between screens and a scene/change is detected.
02 to the quantizer 3. The subtracter 2 subtracts the input moving image signal and the prediction signal supplied from the frame memory 5, and supplies a difference signal to the orthogonal transformer 202 as an output. Orthogonal transformer 202 performs orthogonal transform such as discrete cosine transform or Hadamard transform on the difference signal and supplies transform coefficients to quantizer 3 via line 201 .

The quantizer 3 quantizes the transform coefficients supplied from the orthogonal transformer 202, but when a scene change occurs and a scene change signal is given from the scene change detector 1, the quantizer 3 quantizes the transform coefficients supplied from the orthogonal transformer 202. In the case of using the enlarging method, the dead zone is enlarged as shown in A in FIG. 4, and after a while, the dead zone is slightly reduced to become as shown in B in FIG. Then, after an additional period of time, the dead zone is returned to its normal value. The same quantizer as in the above embodiment can be used in the case of a method of enlarging the dead sea 7, a method of coarsening the quantization step, and a method of combining the above two methods. The output of the quantizer 3 is supplied to the orthogonal inverse transformer 203 and the variable length encoder 6. The orthogonal inverse transformer 203 inversely transforms the quantized transform coefficients by orthogonal inverse transform corresponding to the previously performed orthogonal transform, and supplies the reproduced difference signal to the adder 4 . The adder 4 reproduces the locally decoded signal using the difference signal supplied from the orthogonal inverse transformer 203 and the prediction signal supplied from the frame memory 5. The locally decoded signal of the output of adder 4 is:
It is supplied to frame memory 5. frame memory 5
is 5. The locally decoded signal is delayed by one frame time and the predicted signal is supplied to the subtracter 2 and adder 4 as an output. The variable length encoder 6 performs variable length encoding on the transform coefficients supplied from the quantizer 3 using an efficient code such as a no-fuman code, and matches the encoding speed with the transmission path speed. 6
The compressed variable length code is supplied to the transmission path via 01. In the decoder, an orthogonal inverse transformer having the same configuration as the orthogonal inverse transformer 203 is added between the variable length decoder 7 and the adder 8 in FIG. The variable length code supplied from the transmission path is variable length decoded while matching the speed of the transmission path and the decoding speed, and the decoded transform coefficients are supplied to the orthogonal inverse transformer. The orthogonal inverse transformer reproduces the original difference signal from the transform coefficients supplied from the variable length decoder 7 and supplies it to the adder 8 .

Thereafter, the adder 8 and frame memory 9 operate in the same manner as in the previous embodiment.

(Effects of the Invention) As explained in detail above, in the present invention, immediately after a sea/change occurs, the dead zone or step of the quantizer is expanded, and after a while, the dead zone or step is slightly reduced. , and then perform normal quantization after a further delay. In this way, immediately after a scene change, the dead zone or step of the quantizer is expanded, and gradually reduced over time.
By suppressing the generation of excessive information due to a non-zero amplitude difference at the time of a change and preventing the encoding from stopping, smooth encoding without discontinuous movements can be performed.

As described above, when the present invention is put into practical use, the effect of improving image quality in video encoding is extremely large.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the basic configuration of an apparatus for implementing the image signal encoding method of the present application, FIG. 2 is a block diagram showing the basic configuration of an apparatus for implementing the conventional image signal encoding method, Figure 3 is a diagram showing the frame chain of the image signal, Figure 4
The figure shows the dead zone of the quantizer in the present invention, Figures 5 (A) to (C) show the quantization steps in the present invention, and Figure 6 shows the dead zone of the quantizer in the present invention. A block diagram showing the embodiment, FIG. 7 is a diagram showing a decoder that reproduces the original moving image signal from the variable length code output from the embodiment in FIG. 6, and FIG. 8 is a diagram showing quantization in the embodiment in FIG. 6. FIG. 9 is a block diagram showing one embodiment of the third invention of the present application. l...Scene change detector, 2...Subtractor, 3.
... Quantizer, 4... Adder, 5... Frame memory, 6... Variable length encoder, 7... Variable length decoder,
8... Adder, 9... Frame memory, 202.
... Orthogonal transformer, 203 ... Orthogonal inverse transformer, 301.
...ROM-P. 302...ROM-N, 3Q3...ROM-M, 3
04... Control circuit, 305... Timer, 306.
...Switcher. Agent Patent Attorney Shinsuke Honjo Figure 3

Claims (3)

[Claims] (1) When performing predictive encoding or transform encoding of a video signal, the change in image content between screens is evaluated, and if it is determined that the image content has changed significantly between screens, a prediction error is detected. Alternatively, an image characterized by expanding at least one of the dead zone or step of the quantization characteristic for the transform coefficient, and gradually reducing it over time to return to the predetermined original dead zone or step. How the signal is encoded. (2) A device that predictively encodes an input video signal, with means for evaluating whether or not a significant change has occurred in the content of the image between screens, and a difference between the input video signal and the predicted signal. A means for generating a signal, a means for quantizing the difference signal, and when performing the quantization, if the output of the evaluation means indicates that there has been a significant change in screen content, the quantization characteristic is temporarily determined to be dead. quantization control means for enlarging at least one of the zone or step and gradually reducing the dead zone or step of the characteristic over time to return to a predetermined original quantization characteristic; The method further comprises: means for generating a locally decoded signal using the output and the predicted signal; means for generating the predicted signal using the locally decoded signal; and means for variable length encoding the output of the quantizing means. A featured image signal encoding device. (3) An encoding device that encodes an input moving image signal by transform encoding, a means for evaluating whether or not a significant change has occurred in image content between screens; means for generating a difference signal from a predicted signal; means for performing orthogonal transformation on the difference signal; means for quantizing the output of the orthogonal transformation means; If it shows that there has been a significant change in the quantization characteristic, first expand at least one of the dead zone or step of the quantization characteristic, and then gradually reduce the dead zone or step over time to achieve the predetermined original value. quantization control means for restoring the quantization characteristic to the quantization characteristic, means for orthogonally inversely transforming the output of the quantizing means, means for generating a locally decoded signal from the output of the orthogonally inversely transforming means and the prediction signal, and the local decoding An image signal encoding device comprising: means for generating the predicted signal using a signal; and means for variable length encoding the output of the quantization means.