JPS61107887A - Adaptive type forecasting coding system and device for picture signal - Google Patents

Abstract

PURPOSE:To prevent that a demodulating error of a forecasting coding system due to a signal error in the transmission line is hourly continued, in the forecasting coding system changing over plural pieces of forecasting adaptively in accordance with the character of the screen. CONSTITUTION:A reverse converting coding circuit 21 converts a code supplied from the transmission line by a line 121 to a forecasting error signal and a refreshing signal, the refreshing signal supplies to a selecting device 24 and a memory circuit 26 and a forecasting error signal supplies to an adder 22. At a decision circuit 25 of a optimum forecasting system, the forecasting signal and decoding signal decide which forecasting signal is optimum, and the result is supplied to a memory circuit 26. The circuit 26 stores the necessary part of those signals, and the signal necessary for an estimating circuit 27 to estimate the optimum forecasting system is supplied to the circuit 27. The circuit 24 selects one out of forecasting signals in accordance with the signal showing the optimum forecasting method and supplies it to an adder 22, and when the refreshing signal is supplied from a line 130, the forecasting signal based upon the forecasting system determined beforehand from a line 129 is outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to predictive coding of multivalued image signals, and particularly to predictive coding in which a prediction method is adaptively switched according to the characteristics of a screen.

(Prior art and its problems) In predicting multivalued image signals such as television signals, (1) one prediction method is fixedly used, and (2) multiple predictions are fixedly used. Various methods have been considered, such as (3) a case where multiple predictions are used in combination, and (3) a case where a plurality of predictions are adaptively switched according to the characteristics of the screen.

In this third adaptive prediction method, which performs prediction by adaptively switching the prediction method, a block is configured in units of multiple pixels, the optimal prediction method is detected for each block, and the optimal prediction method used is shown. There are two methods: one in which the code and the prediction error at that time are transmitted together, and the other in which the local properties of the image signal are investigated using only information corresponding to already encoded pixels to estimate the optimal prediction method for the next pixel. There is. In the latter case, similar estimation is possible on the receiving side, so there is no need to transmit a code indicating the prediction method used, and the circuit configuration is that simple.The present invention is related to this latter invention. It is. For details of this, for example,
Although it is described in the specification of No. 10899, a brief explanation will be given below.

Although any number of prediction methods may be used in this method, it is assumed that, for example, two prediction methods (1) and (2) are used. For ■ and ■, intra-frame prediction method, inter-frame prediction method, etc. can be considered. Let the time on the scanning line (line) on the screen of the pixel signal that has just been manually input be defined as i, and by that time the encoding and Figure 1(a) shows which prediction method has the smaller absolute value of prediction error for decoded pixels, that is, the table showing the optimal prediction method. This is called the reference pattern of the prediction method. From this reference pattern, it is immediately clear that at time i+1, the absolute value of the prediction error will be smaller in prediction method (2). Let's generalize this. When considering a method of referring to M encoded/decoded pixels in order to estimate the optimal prediction method using n types of prediction methods, n8 states can be assumed for the reference pattern. The optimal prediction method to be estimated for each reference pattern is determined in advance. This reference pattern that corresponds to a prediction method is called a reference table.If this reference table is stored in the encoder/decoder, the encoder/decoder can make the same prediction without transmitting the prediction method used. Since the method can be estimated, the method according to the present invention can be realized.

That is, the encoder/decoder may generate a reference pattern during encoding/decoding, and then check the reference pattern against a reference table to determine which prediction method to use for encoding/decoding. At this time, it would be wasteful to generate a reference pattern from scratch for each pixel, so memorize several lines of which prediction method was optimal for each pixel and recall the amount necessary to create a reference pattern for each pixel. It is common practice to create a reference pattern and estimate an optimal prediction method based on the above two methods.However, the problem is what to do when a signal error occurs in the transmission path. In other words, in this method, instead of transmitting a signal indicating the prediction method, the transmitting side uses only the information corresponding to the encoded pixels to determine the prediction method, and the receiving side only uses the information corresponding to the decoded pixels. If the information corresponding to encoded pixels and the information corresponding to decoded pixels do not match due to a signal error in the transmission path, the prediction method is and the receiving side, and an error occurs when decoding the pixel signal. Further, since this erroneously decoded pixel signal is used for estimating a prediction method when decoding subsequent pixels, decoding errors continue over time.

According to the prior art, as a measure against decoding errors in predictive coding methods caused by signal errors in the transmission path, it has been common to rewrite the screen, which is called refresh. In other words, when the amount of code transmission is available, decoding errors continue over time by intermittently and forcibly using an encoding method that does not use information corresponding to decoded pixels on the receiving side. This is to prevent However, in the method related to the present invention, it is not possible to prevent decoding errors from continuing over time, because as mentioned above, in the method related to the present method, in order to create a reference pattern, coded and decoded There is a memory/circuit that stores which prediction method was optimal for pixels that have already been converted, but incorrect information may be stored in that memory circuit due to signal errors in the transmission path. 0 For example, in Figure 1, the encoder side stores the optimal prediction method for encoded pixels as shown in the table (a), but an erroneous image signal is generated due to a signal error in the transmission path. As a result of the decoding, the optimal prediction method is determined and stored on the decoder side as shown in the table (b). Then, from the table in (a), pixel time i+
1, prediction method ■ is estimated to be optimal (
From the table in b), it is estimated that prediction method ■ is optimal at pixel time 1+1, and different predictions will be used in the encoder and decoder, resulting in erroneous decoding. Errors continue over time because it is determined which prediction method is optimal for a given image signal, stored, and used to create a reference pattern for a subsequent image signal. However, no method of correcting such errors has been considered in the prior art.

(Object of the Invention) An object of the present invention is to reduce decoding errors in the predictive coding method due to signal errors in the transmission path in a predictive coding method that adaptively switches a plurality of predictions according to the characteristics of the screen. The purpose is to provide a method to prevent temporal continuation.

(Configuration of the Invention) In the present invention, on the encoder side, there is provided a differentiator that generates a difference signal between a predicted signal and a human input signal, a quantizer that quantizes the difference signal, and a quantizer that quantizes the difference signal. an adder that generates a locally decoded signal by adding the predicted signal and the aforementioned predicted signal; a predictor that stores the locally decoded signal and generates a plurality of predicted signals; Generated based on the code conversion circuit that encodes the refresh signal that indicates execution/stop of refresh performed for the purpose of preventing the decoding error caused by it from continuing over time, and the amount of code generated in the code conversion circuit. A refresh determination circuit that generates a refresh signal at regular intervals when there is enough code, and a refresh judgment circuit that generates a refresh signal at regular intervals, and calculates the optimum value for that pixel from multiple prediction signals output from the aforementioned predictor and a locally decoded signal generated by the aforementioned adder. A circuit that determines the prediction method that was
Information called a reference pattern is created from a circuit that stores the result and writes a specific signal only when it receives the refresh signal mentioned above, and an optimal prediction method that corresponds to the encoded pixel signal stored there. an estimation circuit that estimates an optimal prediction method for generating a subsequent prediction signal;
From the estimation result, one of the multiple prediction signals generated by the above-mentioned predictor is extracted as the above-mentioned prediction signal and is supplied to the above-mentioned differencer and adder. However, when receiving the above-mentioned refresh signal, The decoder side has a selector with a function of supplying a prediction signal based on the prediction method.The decoder side also has an inverse transform encoding circuit that converts the transmitted code and generates a prediction error signal and a refresh signal, and its prediction. An adder that generates a decoded signal by adding an error signal and a predicted signal, a predictor that stores the decoded signal and generates multiple predicted signals, and multiple predicted signals output from the predictor. and a circuit that determines the optimal prediction method for that pixel from the decoded signal obtained by the above-mentioned adder, and stores the result as necessary, but only when the above-mentioned refresh signal is received, a specific signal is output. and an estimation circuit that creates information called a reference pattern from the optimal prediction method corresponding to the encoded pixel signal stored therein and estimates the optimal prediction method for generating the next prediction signal. From the estimation result, one of the plurality of prediction signals generated by the aforementioned predictor is extracted and supplied to the aforementioned differencer and adder, but when the aforementioned refresh signal is received, the prediction signal is generated based on a predetermined prediction method. It has a selector that has the function of supplying a predicted signal.

(Principle of the Invention) Since refreshing involves the generation of a large amount of code, it is better not to perform refresh when the amount of generated codes in the code conversion circuit is full of the transmission capacity.When the amount of generated codes is sufficient for the transmission capacity, refresh is performed at regular intervals Therefore, the code conversion circuit has a function of generating a signal indicating the amount of generated code, and using this signal, the refresh determination circuit generates a refresh signal.

For refreshing, a method using encoding that does not use a decoded image signal, such as PCM encoding, can be considered. To do this, a selector that extracts a signal estimated to be optimal from among a plurality of predicted signals may be provided with a function of outputting zero without using any predicted signal when receiving a refresh signal. At this time, the refresh signal can also be supplied to the quantizer to switch to one suitable for PCM encoding with quantization characteristics.

Alternatively, for a certain period, for example, one line, only the prediction signal for the first pixel may be determined in advance, and subsequent pixels may be encoded using previous value prediction. In this way, that one line can be encoded without regard to the previously encoded image signal, and the amount of generated code can be much reduced compared to PCM encoding.

During this refresh period, the refresh signal is used to refresh the memory circuit used in the explanation of FIG. 1 that stores the optimal prediction method for encoded pixel signals. may be done separately from the former refresh, but it is easier to handle if done at the same time. Refreshing of this memory circuit can be done by writing a predetermined signal instead of writing the signal indicating the optimal prediction method supplied from the optimal prediction method determination circuit described above. It may be specially provided and written in. By doing this, even if the reference patterns created by the estimation circuit of the encoder/decoder have been different up to that point, they will always match at this stage.

(Embodiment of the present invention) An example of an encoder configured according to the present invention is shown in FIG. The difference between the original signal input from the line 101 and the predicted signal supplied from the line 110 is calculated by a subtractor 1 and outputted from the line 102 as a prediction error signal.The signal is quantized by the quantizer 2 and then converted to the line 1.
The signal is supplied from 03 to the adder 3 and the code conversion circuit 1O. Adder 3 generates a locally decoded signal by adding the quantized prediction error signal supplied from line 103 and the prediction signal supplied from line 110, and outputs the locally decoded signal to predictor 4 through line 104.
and is supplied to the optimal prediction method determination circuit 6.

The predictor 4 stores as many locally decoded signals as necessary, generates a plurality of prediction signals for the next image signal, two in this example, and outputs them through lines 105 and 106. The optimal prediction method determination circuit 6 determines which prediction signal is optimal from these prediction signals and the locally decoded signal supplied from the line 104, and sends the result through the line 107 to the storage circuit 7.
The storage circuit 7 stores the necessary amount of these signals and supplies the signals necessary for the estimation circuit 8 to estimate the optimal prediction method to the estimation circuit 8 through the line 108.
When a refresh signal is supplied through 12, instead of storing the prediction method, a signal indicating refresh is stored. Using the optimal prediction method corresponding to the encoded signal supplied from line 108, estimating circuit 8 estimates the optimal prediction method and supplies it to selector 5 through line 109. The selector 5 selects one of the prediction signals supplied through the lines 105 and 106 according to the signal indicating the optimal prediction method supplied from the line 109, and sends it to the differencer l and the adder 3 through the line 110.
However, when a refresh signal is supplied from line 112, a prediction signal based on a predetermined prediction method is output from line 110. The code conversion circuit 10 is connected to line 1
The refresh determination circuit 9 encodes the prediction error signal supplied through the line 03 and the refresh signal supplied through the line 112 and sends it to the transmission path through the line 113, and supplies the generated code amount to the refresh determination circuit 9 through the line 111. When the amount of generated code supplied through line 111 is sufficient, a refresh signal is generated at regular intervals and is supplied through line 112 to code conversion circuit 10, selector 5, and storage circuit 7.

FIG. 3 shows an example of a decoder constructed according to the present invention. The inverse conversion encoding circuit 21 converts the code supplied from the transmission line through the line 121 into a prediction error signal and a refresh signal.The refresh signal is supplied to the selector 24 and the storage circuit 26 through the line 130, and the prediction error signal is transmitted through the line 130. 122 to the adder 22. Adder 22 adds the prediction error signal and the prediction signal supplied through line 129 to generate a decoded signal, outputs it through line 123, and supplies it to predictor 23 and optimal prediction method determination circuit 25. The predictor 23 stores the decoded signals as necessary and generates a plurality of, in this example two, prediction signals corresponding to the next signals and outputs them through lines 124 and 125. The optimal prediction method determination circuit 25 determines which prediction signal is optimal based on these prediction signals and the decoded signal supplied from the line 123, and supplies the result to the storage circuit 26 through the line 126. The necessary amount of these signals is stored and the estimating circuit 27 supplies the signals necessary for estimating the optimal prediction method to the estimating circuit 27 through a line 127, and when a refresh signal is supplied through a line 130, the prediction method Instead of storing , a signal indicating refresh is stored. Further, the estimation circuit 27 estimates the optimal prediction method using the optimal prediction method corresponding to the encoded signal supplied from the line 127, and supplies the estimated optimal prediction method to the selector 24 through the line 128.

The selector 24 supplies one of the prediction signals supplied through lines 124 and 125 to the adder 22 through a selection line 129 in accordance with a signal indicating the optimal prediction method supplied from line 128, but a refresh signal is supplied from line 130. When supplied, a prediction signal based on a predetermined prediction method is output from line 129.

(Effects of the Invention) According to the present invention, it is possible to prevent decoding errors from occurring continuously over time due to errors in the transmission path in a predictive coding method that uses multiple prediction methods. . Furthermore, even if error correction coding is applied to the transmitted signal, if the error rate of the transmission path exceeds its correction ability, the adaptive prediction method used as an example of the present invention may be used unless the present invention is applied. cannot be used.

[Brief explanation of drawings]

1(a) and 1(b) are diagrams showing the principle of the present invention, FIG. 2 is a block diagram of an encoder constructed according to the present invention, and FIG.
The figure is a block diagram of a decoder constructed in accordance with the present invention. In Figures 2 and 3, 1 is a differencer, 2 is a quantizer, and 3
．． 22 is an adder, 4.23 is a predictor, 5.24 is a selector, 6.25 is an optimal prediction method determination circuit, 7.26 is a storage circuit, 8.27 is an estimation circuit, 9 is a refresh determination circuit, 10 is code conversion port 1 Figure (b)

Claims (3)

[Claims] (1) It has means for determining and storing which prediction method is the best prediction for encoded pixels among a plurality of prediction methods, and among the stored information, at the current time It has means for estimating the optimal prediction method at the current time from information stored regarding pixels in the vicinity of the pixel to be encoded (current pixel), and performs predictive encoding. In a predictive coding method that performs similar processing to estimate the optimal prediction method and perform decoding, an intermittently predetermined coding method is used for the purpose of preventing decoding errors due to transmission path errors from continuing over time. Predictive coding characterized in that, at the same time, instead of determining and storing which prediction method is the optimal prediction method, a signal determined in advance on the encoder side and the decoder side is stored. method. (2) A differentiator that generates a difference signal between the predicted signal and the input signal, a quantizer that quantizes the difference signal, and a locally decoded signal that adds the quantized signal and the aforementioned predicted signal. A predictor that stores the locally decoded signal and generates a plurality of predicted signals, and a predictor that stores the locally decoded signal and generates a plurality of predicted signals, and a device that prevents the decoding errors caused by the aforementioned prediction error signal and transmission error from continuing over time. A code conversion circuit that encodes a refresh signal indicating whether to execute or stop the refresh that is intended to be performed, and a refresh signal that is generated at regular intervals based on the amount of generated code in the code conversion circuit when there is a surplus in the amount of generated code. a refresh determination circuit that determines the optimal prediction method for that pixel from a plurality of prediction signals output from the aforementioned predictor and a local decoded signal generated by the aforementioned adder; A circuit that writes a specific signal only when it receives the above-mentioned refresh signal, and is ideal for generating the next predicted signal from the optimal prediction method corresponding to the encoded pixel signal stored there. The estimation circuit estimates a prediction method, and from the estimation result, extracts one prediction signal from the plurality of prediction signals generated by the above-mentioned predictor and supplies it to the above-mentioned difference unit and adder as the above-mentioned prediction signal. An encoding device using a predictive encoding method, characterized by having a selector having a function of supplying a predicted signal based on a predetermined prediction method when a signal is received. (3) an inverse transform encoding circuit that converts the transmitted code and generates a prediction error signal and a refresh signal; an adder that adds the prediction error signal and the prediction signal to generate a decoded signal; A predictor that stores the decoded signal and generates a plurality of prediction signals, and a prediction method that is optimal for that pixel based on the plurality of prediction signals output from the predictor and the decoded signal obtained by the above-mentioned adder. a circuit that stores the results as needed, but writes a specific signal only when receiving the refresh signal mentioned above;
An estimation circuit that creates information called a reference pattern from the optimal prediction method corresponding to the encoded pixel signal stored therein and estimates the optimal prediction method for generating the next prediction signal, and an estimation circuit that estimates the optimal prediction method for generating the next prediction signal, and from the estimation result. One of the multiple prediction signals generated by the aforementioned predictor is extracted and supplied to the aforementioned difference device and adder, but when the aforementioned refresh signal is received, a prediction signal based on a predetermined prediction method is supplied. A predictive coding decoding device characterized by having a selector with functions.