JPH01119185A - Predictive coding system - Google Patents

Abstract

PURPOSE:To cause coding to be high-speed without increasing a processing time in a coding circuit by making the horizontal scanning line of inputted image data into a unit, inputting the data of n-systems obtained after the image data are time-base- expanded to be n-fold into n-sets of two-dimensional predictive coding circuits in parallel, and transmitting and receiving a local decoding value between the circuits mutually. CONSTITUTION:A constitution is so made that the data of n-systems obtained by time-base-expanding the image data to be n (integer equal to two or above)-fold in making the horizontal scanning line of the inputted data into the unit are inputted to n-sets of two dimensional predictive encoding circuits 6-8 in parallel, and that the local decoding value is transmitted and received between the n-sets of two-dimen sional predictive encoding circuits 6-8 mutually. Consequently, a parallel processing can be executed as to the data of n-lines of horizontal scanning lines, a picture element just before can be used for a prediction as to a horizontal direction, and simultaneous ly, by using local decoding data from the two-dimensional predictive coding circuits 6-8 to process the adjoining horizontal scanning lines, even the picture elements of the adjoining horizontal scanning lines can be used for the prediction. Thus, the satisfactory two-dimensional coding can be processed at a high speed.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a predictive coding system, and particularly to a predictive coding system that handles image data.

(Prior Art) As a coding method for digitally transmitting temporally correlated signals such as video signals, predictive coding such as well-known differential PCM coding (hereinafter referred to as DPCM) is known. There is. For example, when transmitting a signal with a large amount of information such as a television signal, it is necessary to reduce the bit rate to a transmission bit rate that is compatible with the transmission path.
One possible method is to use predictive coding.

However, when the transmission rate is extremely high, predictive encoding processing must also be made faster, but there is a limit to how fast this processing can be made. Therefore, the sampled video data is sequentially and cyclically supplied to a plurality of DPCM encoders for each pixel.
A method has been proposed to reduce the CM processing speed to one-fold.

[Problem that the invention seeks to solve]

However, in the above-mentioned method, in order to calculate the predicted value of a certain pixel in each encoder, the data manually inputted immediately before, that is, the data of adjacent pixels is not used. for example,
When raster-transmitted video data is one-dimensionally DPCM encoded, a prediction error becomes large because a predicted value is generated using a pixel that is several pixels ahead in the horizontal direction on the screen. this is,
This is because the correlation between a pixel used to generate a predicted value for a certain pixel and that pixel decreases.

Generally, difference data is nonlinearly quantized to reduce the amount of data, but if the prediction error is large, the difference between the representative value of the nonlinearly quantized data and the true value becomes large, and the transmitted image data deteriorates. Resulting in.

Furthermore, as predictive encoding of images, it is possible to perform so-called two-dimensional predictive encoding that also uses vertical correlation of images, and it is possible to further reduce prediction errors.
When data is processed in parallel using a plurality of predictive encoders, if the processing described above is performed, pixels to be used for prediction cannot be freely selected, and prediction errors cannot be reduced.

In view of the above-mentioned problems, an object of the present invention is to provide a predictive encoding system that can perform extremely effective two-dimensional predictive encoding at high speed without increasing the processing speed within the encoding circuit. .

[Means for solving problems]

For this purpose, the predictive coding system of the present invention expands the time axis of input image data by n times (n is an integer of 2 or more), using the horizontal scanning line of the input data as a unit. n systems of data obtained from the above are input in parallel to n two-dimensional predictive encoding circuits, and locally decoded values are exchanged between the n two-dimensional predictive encoding circuits.

Further, in a preferred embodiment of the present invention, in the above configuration, the processing timing of the n two-dimensional predictive encoding circuits for vertically adjacent pixels of the image is shifted.

(Operation) By configuring as described above, data of n horizontal scanning lines can be processed in parallel, and in the horizontal direction, the immediately preceding pixel can be used for prediction, and adjacent horizontal scanning lines can be processed. By using locally decoded data from the two-dimensional predictive encoding circuit, pixels of adjacent horizontal scanning lines can also be used for prediction, so that good two-dimensional predictive encoding can be processed at high speed.

〔Example〕

An embodiment in which the present invention is applied to a two-dimensional DPCM encoding system for video signals will be described below. Figure 1 (A)
, (B), and (C) are diagrams for explaining a system of a 3fE example of the present invention.

FIG. 1(A) is a diagram showing a schematic configuration of the code section of the system according to the present embodiment, and FIG. 2 is a timing chart for explaining the processing timing of each section in FIG. 1(A). In FIG. 1(A), 1 is a terminal to which a data sequence obtained by sampling a television signal is input, and the input data is distributed to three systems by a data distributor 2.

The data distributor 2 distributes the input data for one horizontal scanning line (H
) are distributed to the line memories 3, 4.5 sequentially and cyclically. FIG. 2(a) schematically shows data input to the distributor 2, and the reference numbers indicate horizontal scanning line numbers. Also, FIG. 2(b).

(c) and (d) show input data to the line memories 3 and 4.5, respectively, diagonal lines indicate no data input, and numbers indicate horizontal scanning line numbers.

Line memories 3 and 4.5 are for expanding the time axis of input data by three times and outputting it. They take in 18 minutes of data and read the data in a period of 3H of input data. conduct. The data output timing of line memory 3.4.5 is as shown in Figure 2 (e), (f), and (g).
A seed is set in which 3H worth of data is output in parallel at the same time. The transmission rate of data transmitted from line memories 3, 4.5 is 1/1/1 of the transmission rate of input data.
3, and these data are input in parallel to two-dimensional DPCM encoders 6, 7.8. As will be detailed later, D
PCM encoders 6, 7.8 perform two-dimensional prediction using data from other encoders 8, 6.7, and supply differential data to line memories 11, 12, 13 in parallel. Here, D
The processing time required for encoding by PCM encoders 6, 7.8 is I
It is assumed that the period is sufficiently short compared to the period of H.

The line memories 11, 12, and 13 take in the differential data, compress the time to ⅓ in units of IH minutes, and output it. This read timing is determined by each memory 11 for each IH period of input data.
, 12 and 13 are set to sequentially output 18 minutes of data.

The differential data read from the line memories 11, 12, and 13 is time-axis multiplexed by the data multiplexing circuit 14, and is output line-sequentially at the timing shown in FIG. The signal is sent out to various transmission paths via the signal.

FIG. 1(B) is a diagram showing the specific configuration of the two-dimensional DPCM encoder 6, 7.8 in FIG. 1(A), and FIG. 1(C) is the encoder of FIG. 1(B). FIG.

First, the internal circuit of the two-dimensional DPCM encoder 6 will be explained. 100 is a terminal to which data read from the line memory 3 is input. 101 is an arithmetic unit that calculates the difference value between the predicted value and the input value, and 102 is a quantizer that nonlinearly quantizes the output of the arithmetic unit 101 using a quantization characteristic Q to reduce the number of bits. Data output from converter 102 is supplied to delay circuit 113 and circuit 103. 103 is a circuit having a characteristic Q −1 opposite to the quantization characteristic Q and setting a representative value for the output value of the quantizer 102;
Reference numeral 104 denotes an adder that adds the representative value and the predicted value of the previous pixel to obtain a locally decoded value. 105 is a predictor, and the output of this predictor 105 is supplied to the arithmetic unit 101 and the adder 104 as a predicted value.

The above components 101 to 105 in the encoder 6 are the same as the components 122 to 126 and 132 to 1 in the encoders 7 and 8.
36 and the predictor 105 of each encoder 6.7.8
, 126, and 136 have the same configuration. Here, explanation will be given taking the predictor 105 as an example.

106 is a delay device that delays the locally decoded value obtained from the adder 104 by a period (D) corresponding to one pixel;
If the pixel for which the predicted value is being calculated is the pixel indicated by S in Figure 1 (C), the output of 6 is the pixel immediately preceding it, that is, the horizontally adjacent pixel (Figure 1 (C) is the local decoding value of (denoted by C). Reference numeral 112 denotes a delay device that delays the locally decoded value output from the adder 135 of the encoder 8 by a period (3H-3D) shorter than three horizontal scanning periods by a period corresponding to two pixels. The three horizontal scanning periods referred to here are shown in Figure 1 (A
), and corresponds to the period during which IH worth of image data is input to the encoders 6, 7.8, respectively. In the encoder 8, a period method (SD) delay device for one pixel is provided on the input side, and as is clear from FIG. The local decoded value of the shifted pixel (shown as A in FIG. 1C) is output from the delay device 112. Further, 109 is an ID delay device, and the delay device 10
The output of 9 is the pixel of the previous line at the same position in the horizontal direction (
(shown as B in FIG. 1C).

The local decoded values of pixel A, pixel B, and pixel C are determined by the coefficient unit 108.
The coefficients a, b, and c are multiplied by 110 and 107, and the adder 1
11 to obtain the predicted value of the pixel S. That is, the first
In order to calculate the predicted value of the pixel S in the figure (C), the local decoded values of the three pixels indicated by circles in the figure are used.

In the encoder 7, the input data is ID-delayed by the delay device 121, so the locally decoded value obtained and output from the predictor 12B is for pixel A. Similarly, in the encoder 8, the input data is delayed by 2D in the delay device 131, so the local decoded value to be predicted and output is for the pixel A.

In this way, due to the presence of the delay devices 121 and 131, the pixels processed by the encoders 6 and 7.8 are shifted in the horizontal direction.
That is, by shifting the processing timing of pixels aligned in the vertical direction of the image in the encoders 6, 7.8, pixels located on both sides in the horizontal direction of the pixel for which the predicted value is to be calculated are used for the calculation. can. Therefore, the output of 7.8 of the prediction error is provided at the same timing for pixels aligned in the vertical direction.

In the above-mentioned encoding system, the transmission rate of the data input to each DPCM encoder 8, 7.8 is 1/3 of that of human data, and the DPCM encoder 6, 7.8 as a whole DPCM encoding can be performed at three times the processing speed of each of the following. In addition, since all pixel data for each horizontal scanning line is sequentially input to each DPCM encoder, encoding can be performed using the correlation between adjacent pixels, and prediction errors can be eliminated when generating predicted values. It doesn't make it a big deal. In addition, since it is possible to use pixels located before and after in the horizontal direction in the immediately preceding line, two-dimensional prediction with high prediction accuracy is possible.

FIG. 3(A) is a diagram showing a schematic configuration of a decoding section corresponding to the code section of FIG. 1, and differential data is inputted line-sequentially to the terminal 21 via a transmission path.

The data distributor 22 sequentially and cyclically supplies these differential data to its IH distribution line memories 23, 24, and 25. The line memories 23, 24 and 25 are configured to expand the time axis by three times and output the 18-minute difference data input during the 18-minute period of manual difference data at the IH minute vehicle position, respectively. As in 3 and 4.5, these are output at the same time. DPCM decoder 26°27.2
8 receives the outputs of the line memories 23, 24, 25 and the decoded values of other decoders, performs DPCM decoding, and supplies the decoded data to the line memories 31, 32, 33. Line memories 31, 32, and 33 each contain 3 input difference data.
18 minutes of decoded data input during a period of H minutes is time-axis compressed to 1/3 in units of IH minutes. Line memory 31.32
．． 33 sequentially outputs 18 minutes of decoded data and inputs it to a data multiplexing circuit 34, which is again multiplexed line-sequentially and outputted from a terminal 35.

FIG. 3(B) is a diagram showing a specific example of the configuration of the decoders 26, 27.28, which correspond to the encoders 6, 7.8 shown in FIG. 1(B). 211° vessel, 201, 212, 21
3 are representative value setting circuits similar to 103, 202 and 21, respectively.
3,223 is an adder that outputs the decoded value, 205,216
, 225 is an arithmetic unit that performs the same operation as the predictor 105, and 20
6 is a delay device (3H-)D) for supplying the decoded value of the scanning line decoded by the decoder 28 in the previous period to the decoder 26; 203 and 214 are the vertical coders of the screen; 26, 27, and 28, the aforementioned encoder 6
, 7.8, the operation of the subtracters 101, 122, 132 and the following is the same, so the explanation will be omitted.

It goes without saying that the above decoding system can perform decoding at three times the processing speed of each DPCM decoder.

In the above embodiment, the predictor 105°126.1
Although the data processing time T from the output to the input of the 36 is ignored, it can no longer be ignored if the processing speed is increased.

Taking into account this time T, the processing timing of the encoder that generates the locally decoded value for use in calculating the predicted value must be preceded by the processing timing of the encoder that calculates the predicted value by T.

Note that the configuration of the two-dimensional predictor is not limited to the configuration of the above-described embodiment. However, it is necessary to design with the existence of the period T in mind.

In addition, in the above embodiment, parallel processing is performed in three systems, but it is generally possible to perform parallel processing in n (≧2) systems, and in this case, the time axis is expanded by n times before the n encoders. Needless to say, a 1/n time-base compression circuit is provided at the subsequent stage of the circuit.

(Effects of the Invention) As explained above, according to the system of the present invention, extremely effective two-way processing can be achieved without increasing the processing speed within the encoding circuit.
It has become possible to perform dimensional predictive coding at high speed.

[Brief explanation of the drawing]

FIG. 1 is a diagram for explaining the code section of a system as an embodiment of the present invention, FIG. 1(A) is a diagram showing its schematic configuration, and FIG. FIG. 1(C) is a pixel arrangement diagram for explaining the operation of the encoder of FIG. 1(B). FIG. 2 is a timing chart showing the operation timing of FIG. 1(A). FIG. 3 is a diagram for explaining the decoding section corresponding to the code section in FIG. 1, and FIG.
FIG. 1(B) is a diagram showing the configuration of a decoder corresponding to the encoder of FIG. 1(B). In the figure, 3.4.5... line memory for time axis extension, 6.
7.8... Encoder for two-dimensional DPCM, 11.12
．． 13... Line memory for time axis compression, respectively, 105.126, 136... Predictor, respectively, 104.1
25, 135...adders that output locally decoded values, 112, 121, 131...delay units for processing timing shift, respectively.

Claims (2)

[Claims] (1) Using the horizontal scanning line of input image data as a unit, expand the time axis of the image data by n times (n is an integer greater than or equal to 2), and convert n systems of data into n two-dimensional predictive codes. A predictive encoding system configured to input local decoded values to the n two-dimensional predictive encoding circuits in parallel and to exchange locally decoded values between the n two-dimensional predictive encoding circuits. (2) The predictive encoding system according to claim (1), wherein the processing timing of the n two-dimensional predictive encoding circuits for pixels aligned in the vertical direction of the image is shifted.