JPS6174486A - Encoding system of picture signal - Google Patents

Abstract

PURPOSE:To make suitably data into a block in terms of data compression by encoding a prediction error with a high prediction hitting ratio in a lump by means of a identifying code. CONSTITUTION:With use of the prediction error of a picture element in the vicinity of an encoded picture element on a scan line having been encode, a signal M for estimating the height of a prediction hitting ratio is formed 7, and continuous sections supposed excellent in prediction hitting ratio are controlled by a detection circuit 8 as a prediction error. Then the coincidence of the prediction is detected 5, and when the prediction in the continuous section with a high prediction hitting ratio is coincident, the section is encoded by a identifying code. When the prediction is dissident, the identifying code and prediction error signal is encoded in terms of a variable length encoder 11. Thus prediction errors with high prediction hitting ratios are encoded in a lump by the identifying code.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image signal encoding method that performs data compression of a multivalued image signal using a predictive encoding method.

[Conventional technology]

A predictive coding method is known as a data compression method for multivalued image signals. This predictive encoding method uses an already input image signal S as a reference image signal, predicts the currently input image signal X, and encodes the prediction error. D P CM (Diff-ere'nential
In the Pulse Code Modulation (Pulse Code Modulation) method, a prediction signal X'' is obtained as a linear combination of reference image signals S.

For example, as shown in FIG. 3, a pixel a adjacent to a pixel to be encoded
, b, and c, in linear prediction, the predicted signal X' is obtained as x'=, a+kz b+, and c. Here, k+, kz.

k□ is a coefficient whose size is the predicted signal X' of each pixel
Indicates the strength of the correlation.

The prediction error signal e is obtained by e=x−x”,
As shown in FIG. 4, the probability of occurrence P (e) of this prediction error signal e has a property that it is maximum at e=0 and concentrates on small values around 0. Therefore, as shown in Figure 5,
1. By assigning a shorter code to the code length L (e), the more likely the prediction error is to occur around 0. Data compression can be performed.

In order to further increase the data compression rate, (1) a means of dividing the prediction error into fixed-length blocks and encoding them (Nakano, Okino: "Date Compression of Gradation Facsimile", Oki Electric Research and Development, Vol. No. 109, Vol, 46.&2, 1972
(Refer to March 2012), and (2) a means of dividing the prediction error into blocks of variable length and encoding them (Omiya: "Coding method for halftone images", Proceedings of the 1981 National Conference of the Institute of Image Electronics Engineers, Floor 3).
8) has been proposed.

The means for m focuses on the fact that pixels with a prediction error of 0 are concentrated, divides the scanning line into blocks of m pixels each (m is a fixed value), and then divides the scanning line into blocks of m pixels each (m is a fixed value), and calculates the It performs mode classification and encoding.

■When the prediction error within one block is O, that block is encoded with only the identification code “0” (skip mode)
.

(2) When a prediction error other than O is included in one block, the identification code "1" is first encoded, and then the prediction errors for m pixels in the block are variable-length encoded (non-skip mode).

An example of encoding by means of (11) is shown in FIG. 6, or the image signal to be encoded indicates the gradation of 16 gradations of pixels. is a prediction error, BL is a block, and CD is a compression code.In block BL2, prediction error e is O, and A, et al.
Compression code CD becomes “0” and blocks BLI, BL3
Then, the identification code "1" and the prediction error 0.1. -1.2
．． Variable length code ε corresponding to -2°... , ε1. ε−
It is composed of compressed code CDs of 1, ε2° ε-2...

In addition, the means (2) above focuses on the strong correlation between scanning lines of a halftone image, and divides the interval in which pixels of the same density continue in the immediately preceding scanning line, which has already been encoded, into 1 pro.7. and is encoded as follows.

(2) When the pixel densities of all encoded scanning lines within a block are the same as the pixel densities of the immediately preceding scanning line, encoding is performed using only the identification code "0" (A mode).

■When the pixel density of the encoded scanning line in a block includes a pixel with a density different from that of the immediately preceding scanning line, the identification code "1" is displayed.
Prediction errors of pixels within a block are each subjected to variable length coding (D mode).

An example of encoding by means of (2) is shown in FIG. 7, where X is the image signal of the scanning line immediately before encoding, X is the image signal of the encoded scanning line, Xo is the predicted signal, and e is a prediction error, BL is a block, CD is a compression code, and block BL1. In BL2, the compression code is "0" because it is in the A mode of (2), and in the block BL3, since it is in the D mode, the compression code is "1ε-1ε.ε.ε1".

(Problem to be solved by the invention) In the above-mentioned conventional encoding method (1), since the block size is fixed, the number of pixels with a prediction error of 0 is m
There is a drawback that an improvement in the compression ratio cannot be expected unless a block contains at least 1 or more consecutive pixels.

Furthermore, in the encoding method according to (2) above, the probability of occurrence of an A-mode program decreases as the size of the program increases, and for this reason, when the program size exceeds a predetermined value, A contrivance has been made to divide the A-mode program into sub-blocks every m pixels, and to encode these sub-blocks using the encoding method according to (1) above. In this way, the length of the pro-noxes is made variable, but if the consecutive length of the same density pixels in the previous scanning line is shorter than the corresponding pixels of the same density in the encoding scanning line, then A However, as in block BL3 in Figure 7, the same-density pixel in the immediately preceding scanning line is even one pixel longer than the same-density pixel in the encoding scanning line. If they were continuous, there was a drawback that it would become D mode. In addition, in areas where the density changes for each pixel in a halftone image, a large number of identification codes are required, so the above-mentioned encoding method (1) has the disadvantage that the amount of codes may increase. Ta.

The present invention aims to improve the above-mentioned conventional drawbacks.

[Means for solving problems]

The image signal encoding method of the present invention is an encoding method that predictively encodes a multilevel image signal, and uses prediction errors of pixels near encoded pixels on an encoded scanning line to increase or decrease the prediction accuracy rate. means for forming a signal for estimating the prediction accuracy; and means for grouping into a block prediction errors in continuous sections in which the prediction accuracy is estimated to be high by the signal formed by the means;
means for detecting a prediction match within an interval, encoding the interval with an identification code of "0" when the predictions match within the interval, and encoding the interval when the predictions do not match within the interval; “
1” identification code and the prediction error of the section are variable-length encoded,
The section in which the prediction accuracy is estimated to be low is equipped with a variable length encoder and a switching means that directly encodes the prediction error into variable length code, and the prediction errors having a high prediction accuracy are collectively coded by an identification code. It is something that becomes.

[Effect]

A signal for estimating the level of prediction accuracy is formed, a continuous section with a high prediction accuracy is defined as one block, and when prediction coincidence is detected by means for detecting prediction coincidence within that section, The interval is encoded with an identification code, and when a prediction mismatch is detected, the identification code and prediction error signal are converted into variable length encoding for that interval, and in the interval where the prediction accuracy is low, the prediction error is used as is. Blocking can be appropriately performed by variable-length encoding and using a signal for estimating the level of prediction accuracy.

〔Example〕

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

FIG. 1 is a block diagram of an embodiment of the present invention, where ■ is an input terminal for an image signal, 2 is a shift register for shifting the image signal, 3 is a predictor, 4 is a subtracter, and 5 is a prediction number. Component circuit, 6 is a shift register, 7 is a mode signal generation circuit, 8
1 is a G-mode continuous detection circuit, 9 is a first-in-first-out memory (FIFO), 10 is a multiplexer (MPX), 11 is a variable length encoder, and 12 is a multiplexer (MPX). ．． 13 is an output terminal.

Assuming that the input image signal is a multivalued image signal (4 bits/pixel) indicating 16 gradations from 0 to 15, the shift register 2 to which this multivalued image signal is input has a configuration of 4×(g+2) words. where a is the number of pixels in the l line. Then, the predicted reference pixel signals a, b, and c of the output signals of the shift register 2 are input to the predictor 3.

The prediction signal X' output from the predictor 3 and the encoded pixel signal X output from the shift register 2 are added to a subtracter 4, and a prediction error signal e is output. This prediction error signal e is obtained from the difference as a value modulo 16.

That is, it becomes. With this operation, the difference went from 33 ways from +15 to -15 to 0 to 1 as shown in Table 1.
5, 16 ways, and the prediction error signal e can be expressed with 4 bits. An example of a variable length code corresponding to the prediction error signal e is also shown, and its code length is also shown.

Table 1 The prediction-number construction circuit 5 checks the prediction error signal e and outputs a signal of "0" when the prediction error signal e is 0. , a signal of 1° is output, and in addition to the shift register 6 and the G mode continuous detection circuit 8, the shift register 6 is also provided with a prediction state indicating whether the prediction error signal e is 0 or not. Signals will be accumulated. The shift register 6 has a configuration of (1 bit)×1 (word) so that 1947 worth of predicted state signals can be stored.

Furthermore, the mode signal generation circuit 7 inputs predicted state signals eb+(+ed) of the pixels d, b, and c of the immediately preceding scanning line, which is the encoded scanning line, from the shift register 6, and calculates all these values. '0'', convert it to G (GO
This is called the B (BAD) mode and outputs "0"; otherwise, it is called the B (BAD) mode and outputs "1". This output signal is the mode signal M, which is a signal indicating the level of the prediction accuracy. This mode signal M is used to block the prediction error signal.

When the mode signal M indicates the B mode, the G mode continuous detection circuit 8 controls the multiplexer 10 to output the prediction error signal e using the control signal i, and also controls the multiplexer 12 to output the prediction error signal e when the mode signal M indicates the B mode. Switching control is performed to selectively output the output signal. In this case, the prediction error signal e is input to the variable length encoder 11, encoded, and output via the multiplexer 12 to the terminal 13.

Further, when the mode signal M indicates the G mode, it is determined whether the prediction error signal e is 0 or not based on the signal from the prediction/number construction circuit 5. This is repeated until the mode signal M indicates B mode, and when it becomes B mode, G
If the prediction error signals e are all O during the period in which the mode is continuous, the multiplexer 12 is controlled by the control signal i to selectively output the identification code "0", and it is written to the memory 9 by the signal j. The entered prediction error signal e is cleared. If the prediction error signal e contains something other than 0 during the continuous G mode, the multiplexer 12 is controlled by the control signal i to selectively output the identification code "1", and then The output signal k of the variable length code circuit 11 is selectively outputted, and at the same time, the multiplexer 10 is controlled to selectively output the output signal of the memory 9, so that the prediction error signal e accumulated in the memory 9 during the G mode period is Multiplexer 1
0 to the variable length encoder 11, and the converted variable length code is outputted via the multiplexer 12.

FIG. 2 is an explanatory diagram of the operation of the encoding method of the present invention, in which the image signal x-2, where X' is a predicted signal indicating the image signal of the immediately preceding scanning line, and the numbers corresponding to each pixel indicate the gradation.

Also, e is a prediction error signal, e'' is a prediction error signal of the previous scanning line, M is a mode signal, BL is a block, and CD is a compression code.

As mentioned above, pixels a, b, and c in FIG. 3 are used as reference pixels for prediction, and in order to form a signal for estimating the prediction accuracy, that is, mode signal M, Prediction error signals e° of pixels S, C, and D are used. Then, when the prediction error signals of pixels S, C, and D all match, the mode signal M is set to "0", making this the G (GOOD) mode; otherwise, the mode signal M is set to "1". This is called B (BAD) mode. Therefore, when the prediction error signal e' is not 0, the three pixels immediately below and on both sides thereof are in B mode. For example, the prediction error signal e“ of block BL3
is -1, so the block BL2 immediately below it and on both sides
．． BL3. The mode signal M of BL4 becomes "1".

As described above, once the mode signal M is formed, it is divided into blocks and encoded as follows.

(When the al mode signal is “O”, the interval in which “0” continues is defined as one block. When the prediction error signal e in this block is all 0, the block is given the identification code “0”.
If the prediction error signal e that is not O is included in the block, the identification code is "1".
The prediction error signal within that block is variable-length coded.

(When the bl mode signal is "1", the prediction error signal is variable-length encoded with one pixel as one block.

For example, in the block BLI in FIG. 2, the mode signal M is an interval of "0" for 3 pixels, and the prediction error signal e is all 0, so the compression code CD is the identification code "0".
”. Also, the next block BL2.BL3.BL
4 is an interval in which the mode signal M is "1", one pixel is defined as one block, and the prediction error signal e is directly converted into a variable length code ε. .

ε. , ε-1. Furthermore, in block BL5, the mode signal M is in an interval of "0" for three pixels, and only one prediction error signal e is l in that interval, so the identification code is "1".
” and variable length codes ε., ε..

ε1.

Next, the operation shown in FIG. 1 when the image signal shown in FIG. 2 is manually inputted will be explained. Shift register 2 to predictor 3
The image signals a, b, and c are inputted to the input terminal and a predicted signal X'' is outputted, and the predicted signal X°, the image signal X of the encoded scanning line, and the input signal 5, xI in FIG.

It is shown in X. The prediction error signal e is as shown in e of FIG. 2, and the prediction error signal e' of the previous scanning line is the second
It is shown at e' in the figure. The mode signal generation circuit 7 includes
The prediction error signal e° for three pixels is input from the shift register 6, and when all of them are "0", the G mode ("0"
”) is output, and the prediction/number construction circuit 5
Then, it detects whether the prediction error signal e is 0 or not, and when the prediction error signal e is 0, it outputs a signal of "0", so after the prediction error signal e' continues to be O, the block BL3
When the value becomes -1 as in
Since e of the shift register 6 becomes "1'," the mode signal M outputs "1" indicating B mode.
With the shift operation in the shift register 6, "1" is sequentially shifted, and even when e6 becomes "1", the mode signal M becomes "l", and the block BL2°B in FIG.
L3. The mode signal M corresponds to BL4.

The G mode continuous detection circuit 8 detects a continuous section in which the mode signal M is "0", and also identifies whether the signal from the prediction/number construction circuit 5 is "0" or not. This is what we are doing. Then, the mode signal M changes from “0” to “1”
During the G mode period when the mode signal M is "0", it is determined whether or not the prediction error signals e are all O based on the signal from the prediction-number output circuit 5. If so, the G mode continuous detection circuit 8 controls the multiplexer 12 using the control signal i to selectively output the identification code "0". Then, the prediction error signal e written in the memory 9 is cleared by the control signal j. That is, the second
As in block B'L1 in the figure, the compression code CD becomes "0".

When the mode signal M becomes "1", the G mode continuous detection circuit 8 controls the multiplexer 10 using the control signal i to selectively output the prediction error signal e, and sends the prediction error signal e to the variable length encoder 11. In addition, the G mode continuous detection circuit 8 controls the multiplexer 12 using the control signal i to convert the output signal k of the variable length encoder 11 into a variable length code.
This is to selectively output. That is, blocks BL2°BL3. BL4 is obtained by converting the prediction error signal e into a variable length code ε6+O+ε−1.

Also, when the mode signal M changes from continuous “0” to “1”,
When identifying that a continuous interval of “0” includes a prediction error signal e other than 0 due to signalling,
The G mode continuous detection circuit 8 controls the multiplexer 12 using the control signal i to selectively output "1", and then switches to the variable length encoder 11 side. Further, the multiplexer 10 is controlled by the control signal i to switch to the memory 9 side, and the prediction error signal e stored in the memory 9 is added to the variable length encoder 11. In FIG. 2, the prediction error signal e of 001 is applied to the variable length encoder 11, and the compressed code CD of block BL5 is "1ε.ε.ε1
” will be outputted via the multiplexer 12.

In the next section in which the mode signal M is "1", the prediction error signal e is variable-length encoded by the variable-length encoder 11 and outputted via the multiplexer 12.

For variable length encoding of the prediction error signal e in the variable length encoder 11, for example, the codes in Table 1 can be used.

The output terminal 13 outputs the compressed code shown in the column of compressed code CD in FIG. It is easy to receive this compression code and decode it into a multivalued image signal, and the first
It is possible to decode by performing the reverse operation of the configuration shown in the figure.

Furthermore, in the above-mentioned embodiments, the case where linear prediction is used is explained, but it is also possible to use non-linear prediction in which the optimum prediction signal is stored for each prediction state and then predicted. Furthermore, it is also possible to use rank prediction using the rank of likelihood of appearance as seen from the predicted state of the encoded pixel as the prediction error. In addition, when forming the mode signal amount, the prediction error of the pixel on the scanning line immediately before the encoded scanning line is used. It is clear that it is also possible to use pixels on a scan line.

〔Effect of the invention〕

As described above, the present invention provides a mode signal generation circuit 7 that forms a mode signal M, that is, a signal for estimating the level of prediction accuracy using prediction errors of pixels near encoded pixels on an encoded scanning line. and a G-mode continuous detection circuit 8 for summarizing prediction errors in continuous sections where the prediction accuracy rate based on the mode signal M is estimated to be high, that is, continuous sections where the mode signal M is "0". and a means for detecting a prediction match such as a prediction match detection circuit 5, when the predictions match within a continuous interval where the prediction accuracy is high, encode that interval with an identification code such as "0". , when the predictions do not match, “1”
”, etc. and the prediction error signal e are variable-length coded, and the section where the prediction accuracy is estimated to be low, that is, the mode signal M
is "1" in the variable length encoder 11. which performs variable length encoding on the prediction error signal e. Since it is equipped with means such as a multiplexer 10, 12, etc., and encodes prediction errors with a high prediction accuracy together using an identification code, it has the advantage of having good encoding efficiency and being able to obtain a large compression ratio. .

[Brief explanation of drawings]

Fig. 1 is a block diagram of an embodiment of the present invention, Fig. 2 is an explanatory diagram of the operation of an embodiment of the present invention, Fig. 3 is an explanatory diagram of pixels on a scanning line, and Fig. 4 is a prediction error signal occurrence probability curve. 5 is a diagram of the assignment curve of the code length to the prediction error signal, and FIGS. 6 and 7 are diagrams for explaining the operation of the conventional example. 1 is an input terminal for image signals, 2 is a shift register, 3. is a predictor, 4 is a subtracter, 5 is a prediction-number construction circuit, 6 is a shift register, 7 is a mode signal generation circuit, 8 is a G mode continuous detection circuit, 9 is a first-in first-out memory ( 10.12 is a multiplexer (MPX), 11 is a variable length encoder, and 13 is an output terminal.

Claims (1)

[Claims] In a coding method for predictively coding a multivalued image signal, means for forming a signal for estimating the level of prediction accuracy using prediction errors of pixels in the vicinity of the coded pixel on the coded scanning line; means for summarizing prediction errors in consecutive intervals in which the prediction accuracy is estimated to be high based on the signal formed by the means; means for detecting prediction coincidence within the interval; and means for detecting prediction coincidence within the interval. encode the section with an identification code when
When the predictions do not match within the interval, the identification code and the prediction error of the interval are variable-length coded, and for the interval where the prediction accuracy is estimated to be low, the prediction error is variable-length coded. 1. An image signal encoding method comprising a variable length encoder and a switching means, and encoding prediction errors with a high prediction accuracy collectively using an identification code.