JPS61245777A - Coder of picture signal - Google Patents

Abstract

PURPOSE:To reduce lots of information quantity by applying multipulse coding to a forecast error signal obtained by forecast-coding a picture signal. CONSTITUTION:The picture signal fed to an input terminal 110 is subject to subtraction with a forecast signal fed from a forecast signal generating circuit 114 by a subtraction circuit 111. The difference, that is, the forecast error signal is fed to a quantizing circuit 112, where the signal is quantized. Thus, the quantized forecast error signal is fed to an inverse quantized circuit 113 and a multi-pulse coding circuit 115. The forecast error signal subject inverse quantization is fed to the forecast signal generating circuit 114, where the forecast signal is generated. The forecast error signal quantized by the multi-pulse coding circuit 115 is subject to multi-pulse coding to obtain a pulse train. The pulse train is fed to a compression coding circuit 116, where the train is subject compression coding and the result is outputted to a transmission line 1010.

Description

DETAILED DESCRIPTION OF THE INVENTION <Field of Industrial Application> The present invention relates to improvements in predictive coding for efficiently coding image signals.

<Prior Art> A predictive coding method is a typical coding method for image signals. In this method, the amount of information to be transmitted is reduced by transmitting the difference between the input signal and the predicted signal, that is, the prediction error signal.The predicted signal is given by a prediction function, and this function type can be modified in various ways. An excellent predictive coding method as shown below can be obtained.

The value of the intraframe predictive coding prediction function is determined only by the signal within the current frame. For example, as shown in FIG. 9, the coordinates (Xs, Y
The value of the prediction function for signal X in a ) is the value of the prediction function immediately before it (
s 1 , Yo ).

In this method, when X belongs to a flat part of the image, that is, a part where the level changes slowly, a is a value very close to X, so the prediction error signal becomes a small value close to OK, and high coding efficiency is obtained. . However, if X belongs to an edge portion of the image, that is, a portion where the level changes steeply, the prediction error signal will not have a large value because it will be a value far from altx, and the encoding efficiency will be low.

The value of the prediction function of the interframe predictive coding method is determined by the signal of the previous frame.For example, as shown in FIG. 10, the value of the prediction function for the signal determined by the signal. In this method, for parts with little movement, b t! Since xK is a very close value, the prediction error signal is a small value close to 0, and the coding efficiency is high. However, for parts of the image with large movements, b is
Since the prediction error signal takes a value far from , the prediction error signal does not have a large value, and the encoding efficiency is low. A frame memory is required to make it into a device.

The value of the motion compensated interframe predictive coding method prediction function is determined by the signal of the previous frame shifted by the motion vector. For example, as shown in FIG. 11, if the motion vector is v = ('vx, ■a), then the value of the prediction function for the signal xtlc at the coordinates (Xs, Ya) is .

Y, -vy). Here, it is the amount of displacement between l frames in an image near the motion vector Vitx, and its direction of magnitude generally differs depending on the position of I. In this method, as long as the motion vector can be determined with high accuracy,
Since the value is very close to c Fix, the prediction error signal is 0.
The smaller the value close to , the higher the coding efficiency will be. 1981
A paper by Kogan et al. K. published in the 2018 IEICE technical report C381-87, “Motion compensated interframe code for conference television signals (according to IJ, information on A to B of interframe predictive coding method according to this method) K, which has been shown to be able to be encoded in terms of quantity, requires a frame memory and a motion vector detection circuit.

<Problems with the prior art> In the conventional example described above, a prediction error signal is obtained for each sample of the image signal, so the number of samples of the prediction error signal is equal to that of the painter signal, and the number is equal to that of the painter signal. be. For example, at a sampling frequency of 10MHz, 3
When sampling an image signal of 0 frames/second,
There are about 330,000 pieces per frame. Even if such a large number of prediction error signals were encoded efficiently, the amount of code would be large and a high compression rate could not be obtained.

<Objective of the Invention> An object of the present invention is to provide an encoding device that significantly reduces the amount of information while suppressing deterioration of image quality compared to conventional predictive encoding of image signals.

<Configuration of the Invention> According to the present invention, there is provided a means for quantizing a prediction error signal, a means for dequantizing the quantized prediction error signal, and a prediction signal based on the dequantized prediction error signal. means for generating the prediction error signal by comparing the prediction signal and the image signal; a means for generating the prediction error signal by comparing the prediction signal with the image signal; and a combination signal obtained by supplying the pulse train to a synthesis filter and the quantized prediction error signal. There is obtained an image signal encoding device characterized by comprising means for determining a pulse train with a small difference and means for compressing and encoding the pulse train as described above.

Further, according to the present invention, there are provided means for generating a predicted signal from an image signal, means for generating a prediction error signal by comparing the image signal and the predicted signal, and a synthesized signal obtained by supplying the pulse train to a synthesis filter. An image signal encoding device is obtained, comprising means for determining a pulse train in which the difference between the prediction error signal and the prediction error signal is small, and means for compressing and encoding the pulse train.

Furthermore, according to the present invention, there is provided a means for determining a pulse train that minimizes the difference between a composite signal obtained by supplying a pulse train to a synthesis filter and a prediction error signal, and generating a prediction signal based on the output of the synthesis filter for this pulse train. and the means to
An image signal encoding device is obtained, comprising means for generating a prediction error signal by comparing the predicted signal and the image signal, and means for compressing and encoding the pulse train.

Principle of the Present Invention The motion image signal of multi-pulse encoding can be considered to be obtained by a combination of certain fundamental waveforms. In addition, when predictive encoding of an image signal is performed, the frequency spectrum of the prediction error signal of the image signal is calculated by multiplying the frequency spectrum of the image signal by the frequency response of the predictive coding circuit, so the prediction error signal of the image signal is also a basic waveform. It can be thought that it was obtained by a combination of Regarding the frequency response of predictive coding circuits, see "Digital Signal Processing of Images" by Fukinuki.
(Nikkan Kogyo Shimbun) It is written in detail on pages 169 to 171. Specifically, as shown in FIG. 12(a), the signal source generation circuit 101 outputs a pulse train (FIG. 12(b)),
12(d) is input to the synthesis filter having an impulse response as shown in the same figure (cj). Then, the output of the synthesis filter 102 has a waveform as shown in FIG. 12(d). The sequence is information that represents the position where an impulse response should be added and its amplitude.The multi-waveform can be approximated using a pulse train and an impulse response.The following describes how to determine these impulse responses and pulse trains. .

FIG. 13 is a diagram showing a method for determining a pulse train. First, a pulse is output from the signal source generation circuit 103 and applied to the synthesis filter 104. The difference between the original signal and the output of the synthesis filter is calculated, and the power calculation circuit 105 calculates the error power. In the signal source generation circuit 103, the position and amplitude of the pulse are varied. Calculate the error power in each state and find the minimum state.

In this way, one optimal pulse is determined.

Next, the pulse determined here is fixed, and the next pulse is determined in the same manner as above. By repeating this operation, a signal similar to the original signal can be synthesized. The characteristics or impulse response of the synthesis filter may be determined in advance, or the optimum one may be selected depending on the signal. When a synthesized signal that is sufficiently close to the original signal is obtained, the characteristics of the pulse train and the synthesis filter are encoded and transmitted.

This system performs multi-pulse encoding on a prediction error signal obtained by predictively encoding an image signal. As the predictive coding, any predictive coding such as intra-frame predictive coding, inter-claim predictive coding, motion compensated inter-frame predictive coding, etc. may be performed. When performing multi-pulse encoding on the prediction error signal of the two-dimensional signal obtained in this way, it is possible to perform multi-pulse encoding on the two-dimensional signal as it is, or by scanning the 211-dimensional signal. 4 Good. The scanning method may be to scan the entire screen in a straight line, such as horizontal scanning, vertical scanning, or diagonal scanning, or to divide the screen into several blocks and perform the same scanning for each block. It's okay.

If the two-dimensional signal is to be multi-pulse encoded as it is, a two-dimensional synthesis filter is used and the pulse train is searched two-dimensionally. Further, when converting into a one-dimensional signal and performing multi-pulse encoding, a one-dimensional synthesis filter is used to perform one-dimensional exploration of the pulse train.

Furthermore, various methods can be considered for pulse train exploration. That is, in the embodiment described below, when searching for a pulse train, a synthesized signal is created each time one pulse is found to find the pulse train. In the paper "Study of multi-pulse driven speech coding method" by Obama et al. K., when searching for a pulse train, the pulse train is obtained without creating a composite signal. It is also possible to obtain a pulse train without creating it. In addition, in the above-mentioned paper ``Study of multi-pulse driven speech coding method'', when searching for a pulse train, the amplitude of the previously determined pulse is fixed and the next pulse is determined.

In the embodiments described below, a case will be described in which this pulse train search method is used. On the other hand, in the paper ``Study of sound source pulse search method in multipulse-driven speech coding method'' by Ono et al. published in the 1983 Proceedings of the Acoustical Society of Japan, No. 157, it is stated that The next pulse is found while adjusting the amplitude of the previous pulse. Although this pulse train search method is not used in the implementation series described below, the pulse train may of course be determined in this manner in the present invention.

The features of the present invention include the following points.

In other words, unlike conventional encoding, which encodes all samples, this method only encodes the pulse train, and the number of pulses in the pulse train is sufficiently small compared to the number of samples of the prediction error signal of the image signal. Therefore, it is possible to significantly reduce the amount of information. The aforementioned paper [Study of Multi-Pulse Drive Type Audio Coding Method] discusses a multi-pulse coding method for audio signals, but in the present invention, multi-pulse coding is applied to image signals.

<Embodiment> Embodiments of the image signal encoding apparatus according to the first to third aspects of the present invention are shown in FIGS. 1 to 3 using block diagrams. All of these are encoding devices that combine predictive encoding and multipulse encoding.

First of all, Figures 1 and 2 show predictive coding of image signals.
This is an embodiment of an encoding device that performs multi-pulse encoding on the prediction error signal obtained as a result.

Of these, FIG. 1 shows a case where a quantization circuit for quantizing a prediction error signal is included in the predictive coding loop, and FIG. 2 shows a case where a quantization circuit is not included.

FIG. 3 shows an embodiment of an encoding apparatus in which a multi-pulse encoding circuit is inserted into the encoding loop in a predictive encoding circuit for image signals.

First, FIG. 1, which is an embodiment of the encoding device according to the first invention, will be described.

The difference between the image signal applied to the input terminal 110 and the prediction signal supplied from the prediction signal generation circuit 114 is calculated by the subtraction circuit 111 . This difference, that is, each prediction error signal is then supplied to the quantization circuit 112 and quantized. Quantization includes linear quantization and non-linear quantization. In the case of linear quantization, the output of the quantization circuit is, for example, a prediction error signal in which the lower bits of the prediction error signal are deleted and the number of bits is reduced. appears. Even in the case of nonlinear quantization, a prediction error signal whose number of bits is reduced according to the nonlinear quantization characteristics appears. The prediction error signal quantized in this manner is supplied to an inverse quantization circuit 113 and a multipulse encoding circuit 115. In the dequantization circuit 113, the quantized prediction error signal is dequantized. In the case of linear inverse quantization, there are also linear inverse quantization and nonlinear inverse quantization.
In the dequantization circuit 113, for example, an operation is performed in which 0 is added to the lower part of the quantized prediction error signal by the number of bits deleted by the quantization circuit to restore the number of bits before quantization. In the case of nonlinear inverse quantization, the operation of returning the number of bits to the number before quantization is performed according to the nonlinear inverse quantization characteristics. It is assumed that the quantization circuit and inverse quantization circuit in the following description of the embodiment operate in the same manner. The dequantized prediction error signal is sent to the prediction signal generation circuit 1.
14 and used to generate a prediction signal.

The multi-pulse encoding circuit 115 performs multi-pulse encoding on the quantized prediction error signal to obtain a pulse train. This pulse train is supplied to a compression encoding circuit 116,
Here, it is compressed and encoded and output to the transmission path 1010.

Further, FIG. 2, which is an embodiment of the encoding device according to the second invention, will be described. The image signal applied to the input terminal 120 is supplied to a predicted signal generation circuit 121 and a subtraction circuit 122. The predicted signal generation circuit 121 generates a predicted signal from the image signal supplied from the input terminal 120 and supplies it to the subtraction circuit 122 . The subtraction circuit 122 calculates the difference between the image signal supplied from the input terminal 120 and the predicted signal supplied from the predicted signal generation circuit 121,
This difference, that is, the prediction error signal, is supplied to the multipulse encoding circuit 123. The multi-pulse encoding circuit 123 performs multi-pulse encoding on the prediction-difference signal to obtain a pulse train. This pulse train is supplied to a compression encoding circuit 124, where it is compression encoded and output to a transmission path 1020.

Next, FIG. 3, which is an embodiment of the encoding device according to the fFJ3 invention, will be described. The difference between the image signal applied to the input terminal 130 and the prediction signal supplied from the pre-1 signal generation circuit 134 is calculated by the subtraction circuit 131 . This difference, that is, the pre-611 error signal, is then supplied to a multi-pulse encoding circuit 132, where it is multi-pulse encoded to obtain a pulse train. This pulse train is supplied to a synthesis filter 133 and a compression encoding circuit 135. The synthesis filter 133 synthesizes a prediction error signal from the pulse train, and this synthesized prediction error signal is sent to the prediction signal generation circuit 1.
34 and used to generate a prediction signal. Furthermore, the pulse train is compressed and encoded in the compression encoding circuit 135 and output to the transmission path 103o.

Each of the embodiments of the encoding apparatus of the present invention shown in FIGS. 1 to 3 has the following advantages. First, in the case of Figure 1, the number of bits of the rounded prediction error signal that quantizes the prediction error signal obtained by predictive encoding is reduced, making it possible to reduce the calculation accuracy of multipulse encoding, and the device configuration is simple. becomes.

Furthermore, the accuracy of FIGS. 2 and 3 is higher than that of FIG. 1.

However, the frequency characteristics of S/N and output noise are different between FIG. 2 and FIG. 3. In other words, in the case of Fig. 2, the frequency characteristics of the noise generated during multi-pulse encoding are transformed by the frequency characteristics of the synthesis filter that synthesizes the prediction error signal from the pulse train on the receiving side, and the high frequency components are reduced. be. The diagram in FIG. 2 is effective for reducing high frequency components.

Next, the case shown in Figure 3 has the advantage that the highest S/N can be obtained, and the frequency characteristics of the noise generated during multi-pulse encoding are not converted by the frequency characteristics of the receiving side synthesis filter, and are unchanged. It has the characteristic that it appears in the output.

Embodiments of Coding Apparatus For each of FIGS. 1 to 3, any predictive coding such as intra-frame predictive coding, inter-frame predictive coding, motion-compensated inter-frame predictive coding, etc. is performed. In multi-pulse encoding, it is possible to perform a two-dimensional search on a pulse train using a two-dimensional filter as a synthesis filter, and a method to perform a one-dimensional search on a pulse train using a one-dimensional filter as a synthesis filter. Various combinations can be considered depending on the characteristics and purpose of the target image signal. For example, there are methods for removing temporal correlation using inter-frame predictive coding or motion-compensated 7-frame inter-frame predictive coding, and removing spatial correlation using two-dimensional multi-pulse coding. If the number is small, a method may be considered in which the horizontal correlation is removed by two-frame intra-predictive coding and the vertical correlation is removed by one-dimensional multi-pulse coding.

Further, FIGS. 1, 2, and 3 show detailed examples of the encoding apparatus in nine cases in which intra-frame prediction, inter-frame prediction, and motion-compensated 7-frame inter-frame prediction are performed, respectively. Continue explaining.

An embodiment of the encoding device in the case where intra-frame prediction is performed for prediction in the first invention of the present application will be described with reference to FIG. The image signal applied to the input terminal 100 is sent to the subtraction circuit 1
0. In the subtraction circuit 10, the input terminal 10
The difference between the image signal supplied from 0 and the prediction signal supplied from the delay circuit 13 is calculated, and this difference, that is, the prediction error signal, is supplied to the quantization circuit 11 which has the function of limiting the number of possible levels. Ru. The prediction error signal quantized in the quantization circuit 11 is sent to a subtraction circuit 16 and an inverse quantization circuit 20.
supplied to The dequantization circuit 20 dequantizes the quantized prediction error signal and supplies it to the addition circuit 12 .

The adder circuit 12 generates a locally decoded signal from the prediction signal supplied from the delay circuit 13 and the inversely quantized prediction error signal. The locally decoded signal is supplied to the delay circuit 13 and used to generate a predicted signal after being delayed by 1 mj elementary time. That is, the output of the delay circuit 13 is supplied to the subtraction circuit 10 and the addition circuit 12.

Next, the pulses generated in the signal source generating circuit 14 are supplied to a synthesis filter 15. The difference between the signal synthesized by the synthesis filter 15 and the quantized prediction error signal supplied from the quantization circuit 11 is calculated in a subtraction circuit 16, and the result is supplied to a power calculation circuit 17. The power calculation circuit 17 calculates the error power between the quantized prediction error signal and the composite signal.

First, in the signal source generating circuit 14, the position and amplitude of the pulses are varied to determine one pulse with the minimum error power. Next, the pulse determined here is fixed and the next pulse is determined in the same manner. By repeating this operation, a pulse train that can be passed through the synthesis filter 15 to synthesize a signal similar to the quantized prediction error signal is determined and sent to the memory 18. The pulse train sent to the memory 18 is supplied to the compression encoding circuit 19, where it is compressed and encoded and output to the transmission line 1000.

Next, an embodiment of a dead-case encoding apparatus that performs interframe prediction for prediction in the first aspect of the invention will be described with reference to FIG. The image signal applied to the input terminal 300 is supplied to the subtraction circuit 30.

The subtraction circuit 30 calculates the difference between the image signal supplied from the input terminal 300 and the prediction signal supplied from the frame memory 33, and this difference, that is, the prediction error signal, has the function of limiting the number of possible levels. The signal is supplied to the quantization circuit 31. The nine prediction error signals quantized in the quantization circuit 31 are supplied to a subtraction circuit 36 and an inverse quantization circuit 40. The dequantization circuit 40 dequantizes the quantized prediction error signal and supplies it to the addition circuit 32 . The adder circuit 32 generates a locally decoded signal from the prediction signal supplied from the 7-frame memory 33 and this quantized prediction error signal. The locally decoded signal is supplied to the frame memory 33 and used to generate a predicted signal after being delayed by one frame time. That is, the output of the frame memory 33 is supplied to a subtraction circuit 30 and an addition circuit 32.

Next, the pulses generated in the signal source generation circuit 34 are supplied to a synthesis filter 35. A difference between the signal synthesized by the synthesis filter 35 and the quantized prediction error signal supplied from the quantization circuit 31 is calculated in a subtraction circuit 36, and the resultant signal is supplied to a power calculation circuit 37. The power calculation circuit 37 calculates the error power between the quantized prediction error signal and the composite signal.

Further, in the signal source generating circuit 34, the position and amplitude of the pulses are varied to determine one pulse with the minimum error power. Next, the pulse determined here is fixed and the next pulse is determined in the same manner. By repeating this reconnaissance, a pulse train that can be passed through the synthesis filter 35 to synthesize a signal similar to the quantized prediction error signal is determined and sent to the memory 38. The pulse train sent to the memory 38 is supplied to the compression encoding circuit 39, where it is compressed and encoded and output to the transmission line 3000.

Next, an embodiment of the encoding device in the case where motion compensated interframe prediction is performed for prediction in the first invention of the present application will be described using FIG. 6.

The image signal applied to the input terminal 500 is supplied to a motion vector detection circuit 50 and a subtraction circuit 51, respectively. The motion vector detection circuit 50 detects the motion of the image using this image signal and the image signal of the previous frame supplied from the frame memory 54 which can store one frame. The motion vector detected by the motion vector detection circuit 50 is output to a variable delay circuit 55 and a compression encoding circuit 61. The subtraction circuit 51 calculates the difference between the l1iii image signal supplied from the input terminal 500 and the motion-compensated prediction signal of the previous frame supplied from the variable delay circuit 55, and this difference, that is, the prediction error signal, is The signal is supplied to a quantization circuit 52 which has the function of limiting the number of levels. The prediction error signal quantized in the quantization circuit 52 is supplied to a subtraction circuit 58 and an inverse quantization circuit 62. The dequantization circuit 62 dequantizes the quantized prediction error signal and supplies it to the addition circuit 53. The adder circuit 53 generates a locally decoded signal from the prediction signal supplied from the variable delay circuit 55 and this quantized prediction error signal. The locally decoded signal is supplied to the frame memory 54 and is used for motion vector detection and prediction signal generation after being delayed by approximately one frame time.

That is, the output of the frame memory 54 is supplied to the motion vector detection circuit 50 and the variable delay circuit 55. The variable delay circuit 55 generates a motion-compensated prediction signal using the motion vector supplied from the motion vector detection circuit 50 and supplies it to the subtraction circuit 51 and the addition circuit 53.

Next, the signal generated by the signal source generation circuit 56 is supplied to a synthesis filter 57. A difference between the signal synthesized by the synthesis filter 57 and the quantized prediction error signal supplied from the quantization circuit 52 is calculated in a subtraction circuit 58, and the result is supplied to a power calculation circuit 59. The power calculation circuit 59 calculates the error power between the quantized prediction error signal and the composite signal.

Further, in the signal source generating circuit 56, the position and amplitude of the pulses are varied to determine one pulse with the minimum error power. Next, fix the pulse found in the step and determine the next pulse in the same way. By repeating this operation, a pulse train that can be passed through the synthesis filter 57 to synthesize a signal similar to the quantized prediction error signal is determined and sent to the input filter 60. The pulse train sent to the memory 60 is supplied to a compression encoding circuit 61. In the compression encoding circuit 61, the motion vector supplied from the motion vector detection circuit 50 and the pulse train supplied from the memory 60 are compressed and encoded. 6000.

Note that as a special case of the above embodiment, a case may be considered in which the doji conversion circuit and the inverse quantization circuit are deleted.

Next, an embodiment of the encoding device in the case where intra-frame prediction is performed for prediction in the second invention of the present application (FIG. 2) will be described using FIG. 7. An image signal applied to an input terminal 800 is supplied to a delay circuit 80 and a subtraction circuit 81. In the delay circuit 80 , the image signal also supplied to the input terminal 800 is delayed by one pixel time to generate a prediction signal, which is then supplied to the subtraction circuit 81 . The subtraction circuit 81 calculates the difference between the image signal supplied from the input terminal 800 and the prediction signal supplied from the delay circuit 80, and this difference, that is, the prediction error signal, is supplied to the subtraction circuit 84.

Next, the pulses generated in the signal source generation circuit 82 are supplied to a synthesis filter 83. The difference between the signal synthesized by the synthesis filter 83 and the prediction error signal supplied from the subtraction circuit 81 is calculated in the subtraction circuit 84, and this difference is supplied to the power calculation circuit 85. The power calculation circuit 85 calculates the error power between the prediction error signal and the composite signal. Further, in the signal source generating circuit 82, the position and amplitude of the pulses are varied to determine one pulse with the minimum error power. Next, the pulse determined here is fixed and the next pulse is determined in the same manner. By repeating this operation, a pulse train that can synthesize the low light signal with the prediction error signal by passing it through the synthesis filter 83 is determined and sent to the memory 86. The pulse train sent to the memory 86 is supplied to a compression encoding circuit 87, where it is compressed and encoded and sent to the transmission path 5ooo.
is output to.

In the second aspect of the present invention, as an embodiment of an encoding apparatus in which inter-frame prediction is performed for prediction, the delay circuit 80 in the encoding apparatus shown in FIG. 7 may be replaced with a 7-frame memory. Further, as an embodiment of the encoding device in which motion compensated interframe prediction is performed for prediction, the delay circuit 80 in the encoding device shown in FIG. 7 is replaced with a frame memory, and motion vector detection is performed in the same manner as in the example shown in FIG. Just add a circuit and a variable delay circuit.

Next, an embodiment of the encoding device in the case where intra-frame prediction is performed for prediction in the third invention of the present application (FIG. 3) will be described with reference to FIG. 8.

The image signal applied to input terminal 700 is supplied to subtraction circuit 70 . The subtraction circuit 70 calculates the difference between the image signal supplied from the input terminal 700 and the prediction signal supplied from the delay circuit 78, and this difference, that is, the prediction error signal, is supplied to the subtraction circuit 73.

Pulses generated in the signal source generation circuit 71 are supplied to a synthesis filter 72. The difference between the signal synthesized by the synthesis filter 72 and the prediction error signal supplied from the subtraction circuit 70 is calculated in the subtraction circuit 73, and the difference is calculated between the signal synthesized by the synthesis filter 72 and the prediction error signal supplied from the subtraction circuit 70.
supplied to The power calculation circuit 74 calculates the error power between the prediction error signal and the composite signal.

Further, in the signal source generating circuit 71, the position and amplitude of the pulses are varied to determine one pulse with the minimum error power.

Next, by fixing the pulse obtained here and returning the next pulse in the same way, a pulse train that can synthesize a signal similar to the prediction error signal by passing it through the synthesis filter 72 is obtained and stored in the memory. Sent to 75. The pulse train sent to the memory 75 is sent to the compression encoding circuit 79 and the synthesis filter 7.
6. A synthesis filter 76 synthesizes a prediction error signal from the pulse train and supplies it to an addition circuit 77 .

Adder circuit 77 generates a locally decoded signal from the prediction error signal supplied from synthesis filter 76 and the prediction signal supplied from delay circuit 78 . The locally decoded signal is supplied to a delay circuit 78 and is used to generate a prediction signal after being delayed by one pixel. That is, the output of the delay circuit 78 is supplied to the subtraction circuit 70 and the addition circuit 77. Compression encoding circuit 79
The pulse train sent to the transmission line 70 is compressed and encoded here.
Output as 00.

Note that when inter-frame prediction or motion-compensated inter-frame prediction is used for prediction, the configuration may be the same as in the embodiments already described.

Here, a device for decoding an image signal encoded by the encoding device of the present invention will be described for reference.

First, FIG. 14 will be explained as an example of a decoding device. The compression-encoded signal supplied from the transmission path 101O is supplied to the expansion decoding north circuit $111.

In the decompression decoding circuit 111 , the compression-encoded pulse train is decompressed and decoded, and then supplied to the synthesis filter 112 . In the synthesis filter 112, the quantized prediction error signals are synthesized and supplied to the inverse quantization circuit 113. In the dequantization circuit 113, the quantized prediction error signal is dequantized.

Inverse quantization includes linear inverse quantization and non-linear inverse quantization. In the case of linear inverse quantization, in the inverse quantization circuit 113, the lower part of the quantized prediction error signal is deleted by the quantization circuit on the transmitting side. An operation is performed in which 0 is added by the number of bits determined, and the number of bits is returned to the number before quantization on the transmitting side. In the case of nonlinear inverse quantization, the operation of returning the number of bits to the number before quantization is performed according to the nonlinear inverse quantization characteristics. It is assumed that the inverse quantization circuit in the following description of the embodiment (IX2X3) also operates in a similar manner. The dequantized prediction error signal is then supplied to the adder circuit 114.

In the addition circuit 4114, the image signal is predictively decoded from the dequantized prediction error signal supplied from the dequantization circuit 113 and the prediction signal supplied from the prediction signal generation circuit 115. The predictively decoded image signal is output to the output terminal 11.
0 and is supplied to the prediction signal generation circuit 115. The predictively decoded image signal read out from the predictive signal generating circuit 115 is supplied to the adding circuit 114 as a predictive signal.

Next, FIG. 15 will be described as an example of a decoding device. The compression-encoded signal supplied through the transmission path 1020 is supplied to the decompression decoding circuit 121 . Decompression decoding circuit 121
, the compression-encoded six-pulse train is decompressed and decoded, and then supplied to the synthesis filter 122. Synthesis filter 122
, the prediction error signals are combined and supplied to the adder circuit 123. In the addition circuit 123, the image signal is pre-i1111f from the prediction error signal supplied from the synthesis filter 122 and the prediction signal supplied from the prediction signal generation circuit 124.
It will be numbered M. The predictively decoded image signal is the output Rokuko 1
20 and the prediction signal generation circuit 124.

The predictively decoded image signal extracted from the predictive signal generating circuit 124 or 121 is supplied to the adding circuit 123 as a predictive signal.

<Effects of the Invention> Unlike conventional predictive coding in which all samples of an image signal are predictively coded, in this method, the prediction error signal obtained by predictively coding the image signal is converted into a multipulse code. Since the number of pulses in the pulse train is sufficiently small compared to the number of samples of the image signal, it is possible to significantly reduce the amount of information.

[Brief explanation of drawings]

1, 2, and 3 are block diagrams for explaining the present invention, FIGS. 4 to 8 are block diagrams showing embodiments of the present invention, and FIGS. 9, 10, and 11 are block diagrams for explaining the present invention. FIG. 12 is a diagram for explaining a conventional method. FIG. 13 is a diagram for explaining the multi-pulse encoding method, and FIGS. 14 and 15 are block diagrams showing an example of the configuration of a decoding device related to the present invention. In the figure, 10.16, 30, 36, 51.58.70, 73.8
1,84°111.121.131-・Subtraction circuit, 11
, 31.52°112-・ Quantization circuit, 12, 32.
53.77・-addition circuit, 13.78.80-delay circuit, 14,34.56,71°82.101,103-・
Signal source generation circuit, 15, 35.57° 72.76.83
, 102, 104, 133 - synthesis filter, 17.37
．． 59.74.85.105--Power calculation circuit, 18.
38.60.75.86--Memory, 19,39,6
1゜79.87,116,124,135--compression encoding circuit, 20.40,62.113--inverse quantization circuit,
33.54--Frame memory, 50... Motion vector detection circuit, 55... Variable delay circuit, 114, 122,
134-Prediction signal generation circuit, 115, 123.132-
The multi-pulse encoding circuits are respectively shown.

Claims (3)

[Claims] (1) means for quantizing a prediction error signal; means for dequantizing the quantized prediction error signal; and means for generating a prediction signal based on the dequantized prediction error signal; means for generating the prediction error signal by comparing a prediction signal and an image signal; and means for generating the prediction error signal by supplying the pulse train to a synthesis filter to generate a pulse train in which the difference between the synthesized signal obtained by supplying the pulse train to a synthesis filter and the quantized prediction error signal is small. An image signal encoding device comprising: means for determining the pulse train; and means for compressing and encoding the pulse train. (2) means for generating a prediction signal from an image signal; means for generating a prediction error signal by comparing the image signal and the prediction signal; and a composite signal obtained by supplying a pulse train to a synthesis filter and the prediction. An image signal encoding device comprising: means for determining a pulse train having a small difference from an error signal; and means for compressing and encoding the pulse train. (3) means for supplying the pulse train to a synthesis filter to obtain a pulse train in which the difference between the resulting composite signal and the prediction error signal is small; and means for generating a prediction signal based on the output of the synthesis filter for this pulse train; means for generating the prediction error signal by comparing the prediction signal and the image signal;
An image signal encoding device comprising: means for compressing and encoding the pulse train.