JPH0214830B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an image encoding device for highly efficient transmission of halftone image signals.

Conventionally, this type of image encoding device has been equipped with a prediction ranking table on the transmitting and receiving sides that inputs the value taken by the reference pixel and outputs the predicted ranking of the value that the pixel of interest should take, and uses the signal level of the pixel of interest as a predicted ranking value. There is a known device that converts the data into code and transmits the coded data.

According to information theory, if the probability of appearance of each symbol a ₀ , ..., a _o-1 is P ₀ , ..., P _o-1 , the amount of information H per symbol is H= _oi −〓 ⁱ⁼⁰ pilog It is given in ₂ pi (bits). H bits per pixel is the reduction limit in information theory, and this is achieved when the codeword length l _i of symbol a _i can be set as l _i =-log ₂ p _i (bit). be.

FIG. 1 is a diagram showing the relative positional relationship between the pixel of interest X and neighboring pixels A, B, C, and D that are of high reference value, when the pixel of interest is X.

In the figure, the level taken by one pixel is 0 to 15.
When two pixels A and B are selected as reference pixels, there are 16 patterns, and FIG. 2 shows the frequency order of the density level value of the target pixel X with respect to the density level values of A and B, and part of the probability thereof.

From Figure 2, for example, the concentration level is A=0, B=
In the state of 0, the probability that the predicted ranking is 0 and the concentration level of X is 0 is approximately 91%. On the other hand, when the concentration level is A=2 and B=8, the probability that the predicted ranking is 0 is about 25%.

Such predictive coding methods have the following problems.

(1) Since the probability of appearance of each prediction rank differs depending on the state of the reference pixel, it is desirable to use a ^different code depending on the state. Let the number be n
Since there are 2 ^m×n , if we use different codes for each state, for example, the concentrations of A and B in Figure 2 are
2 ^{m × n} combinations must be considered, making the device extremely complex.

(2) Also, prediction ranks 0 to 15 are expressed in 4 bits, but considering that the higher prediction ranks have a high probability of appearance, when encoding the prediction ranks, the upper ranks are expressed in 1 to 2 bits. It can be said that a method in which lower ranks are encoded using a larger number of bits is superior to a method in which all ranks are encoded using the same number of bits.

The present invention solves the problems (1) of the conventional ones as mentioned above.
This was done in view of (2), and for problem (1), it focuses on the fact that the probability of each predicted rank value of the pixel of interest differs depending on the difference in density level between multiple reference pixels. Considering a smaller number of modes than the number of possible states of the reference pixel, both the sending and receiving sides have information on the mode to which each state belongs, and encodes the predicted rank value for each mode using a code suitable for that mode. I decided to do so. To solve problem (2), we decided to encode multiple bits indicating the prediction order at once. Thereby, it is an object of the present invention to provide an image encoding device that can perform encoding according to the characteristics of an information source and can perform transmission in a short time.

Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

Examining the probability of appearance of each predicted rank value in each state, when the values taken by several reference pixels adjacent to the pixel of interest are the same or the level difference is small, it corresponds to the flat part of the image, Here, the correlation between pixels is strong and the entropy per pixel is relatively small. On the other hand, when the values taken by reference pixels vary, the correlation between pixels is weak and the entropy per pixel is relatively large.

Therefore, in one embodiment of the present invention, the signal levels of pixel B directly above the pixel of interest and pixel A immediately to the left are compared, and if the difference is 0 or a relatively small value, the mode is set to S (Strong; When this is not the case, the W (weak; correlation is weak) mode is defined. The level difference between these two modes is related to the number of levels 2 ^m of the original signal, and is set to 0 if m=4, and ±3 if m=7, for example. In other embodiments of the present invention, the mode is determined by taking into consideration the pixel D diagonally to the upper right of the pixel of interest, but this is related to the sampling density of the image, and the sampling density is A fine density indicates that pixel D also needs to be taken into account.

Figure 3 shows an example of the appearance probability of each rank value for both modes when the sampling density is 8 x 8 (pieces/mm), m = 4, and A = B = D is set to S mode, and the others are set to W mode. ,
In this case, if a different code is applied to each mode, the entropy can be reduced by about 10% compared to the case where the two modes are not separated.

When separating into both modes in this way, the transmission format of the encoded signal becomes a problem, but if run-length encoding is not performed first, there is no need for a signal to indicate the mode, and the receiving side In this case, each time each pixel is restored, the mode to which the next pixel belongs is known, and therefore the code used is also known, making it possible to decode the next code word. In addition, when using run-length encoding, multiple rank values of one mode are encoded simultaneously, but if decoded pixels are stored in multiple long registers, the states of surrounding pixels can be The mode can be determined from the above, and no special signal is required for differentiation. On the receiving side, by storing a plurality of decoded ranking values in a register or the like and setting them appropriately, each pixel can be sequentially restored, and since the mode to which the next pixel belongs can also be known, code words can always be decoded.

FIG. 4 shows a block diagram of the transmitting and receiving side including the image encoding apparatus according to the present invention. In the figure, 41
is a prediction converter that stores a prediction conversion table, and converts the density level signal of the pixel of interest X into a prediction rank value C. Reference numeral 42 denotes a mode signal generator, which generates a signal M indicating the mode to which the predicted ranking value belongs in accordance with the image signal A and B density patterns output from the pixel memory 43. 45 is an S mode encoder;
46 is a W-mode encoder, and the prediction rank value C is inputted to one of them according to the mode signal M. Reference numeral 47 denotes a transmission code word selector which controls the transmission order of code words in both modes in accordance with the mode signal M so that the receiving side can decode the data. Reference numeral 51 denotes a received signal buffer, which operates an S-mode decoder 52 and a W-mode decoder 53 according to a mode instruction signal M' that is sequentially input.
Send the received signal to either. 54 is a switch,
The decoded prediction rank value C' is sent to the prediction inverse transformer 55, and the reproduced signal X' is obtained by the inverse transformer 55.
57 is a pixel memory, which contains already restored pixel signals.
A' and B' are sent to a predictive inverse transformer 55 and a mode signal generator 56. Both mode decoders 52 and 53 have registers for holding the decoded prediction ranking values, so that decoding is possible even in the case of run-length encoding.

Next, the operation will be explained.

In this apparatus configured as described above, the prediction converter 41 converts the density level signal of the pixel of interest X into a prediction rank value C according to the density level signals of the reference pixels A and B, and the mode signal generator 42 Both pixels A above,
A mode signal M indicating the mode to which the predicted ranking value C determined by the density pattern of B belongs is output. Then, the switch 44 selects the predicted rank value C based on the mode signal M.
is input to either the S mode encoder 45 or the W mode encoder 46, and the code word selector 47 sends the output of one of the mode encoders 45 and 46 to the transmission line 100.

On the other hand, on the receiving side, the mode signal generator 56 generates the mode signal M' according to the pixel signals A' and B' which have already been decoded and inversely transformed, and the receiving signal buffer 51 generates the receiving signal according to the mode instruction signal M'. It is sent to either the S mode decoder 52 or the W mode decoder 53. The switch 54 sends the decoded prediction rank value C' to the prediction inverse transformer 55, and the inverse transformer 55 obtains a reproduced signal X'.

As described above, the present invention focuses on the fact that the probability of each predicted rank value of the pixel of interest differs depending on the reference pixel pattern, and considers a smaller number of modes than the number of possible states of the reference pixel, and each state Information on the mode to which the prediction rank belongs is held on both the transmitting and receiving sides, and the predicted rank value is encoded for each mode using a code suitable for that mode, and multiple bits indicating the predicted rank value are encoded at once. Therefore, it is possible to perform encoding according to the characteristics of the information source, and the transmission time can be reduced.

[Brief explanation of drawings]

Figure 1 is a diagram showing the pixel arrangement, Figure 2 is a diagram showing a part of the table showing the values of X arranged in descending order of prediction order with respect to the values taken by reference pixels and their probabilities. FIG. 3 is a diagram showing the appearance probability of each predicted ranking value in both modes, and FIG. 4 is a block diagram of an image encoding apparatus according to an embodiment of the present invention. 41 is a prediction converter, 42 is a mode signal generator,
43 is a pixel memory, 44 is a switch, 45 and 46 are S mode encoders, W mode encoders (plural encoders), 47 is a transmission code word selector, 51 is a received signal buffer, and 52 is an S mode decoder. , 53 is a W mode decoder, 54 is a switch, 55 is a predictive inverse transformer,
56 is a mode signal generator, 57 is a pixel memory,
is the pixel of interest, and A, B, and D are reference pixels. In the figures, the same reference numerals indicate the same or corresponding parts.

Claims (1)

[Scope of Claims] 1. A prediction ranking table that defines a predicted ranking value of a pixel of interest with respect to density level signals of a plurality of reference pixels and a density level signal of a pixel of interest, and according to the contents of the prediction ranking table. A prediction converter that converts the density level state of the pixel of interest into a predicted rank value of multiple bits; and a mode that generates a different mode signal depending on whether the difference in density level between the plurality of reference pixels is less than or equal to a certain value. a signal generator, and a mode provided corresponding to each mode for collectively encoding the plurality of bits indicating the predicted rank value using a code suitable for each mode according to the mode indicated by the output of the mode signal generator. Image encoding, comprising a plurality of encoders and a transmission code word selector that selects one of the encoded outputs of the plurality of encoders and sends it to a transmission path in accordance with the mode signal. Device. 2. The image encoding device according to claim 1, wherein the plurality of reference pixels are a pixel to the left of the pixel of interest and a pixel directly above the pixel of interest. 3. The image according to claim 1, wherein the plurality of reference pixels are a pixel to the left of the pixel of interest, a pixel directly above the pixel of interest, and a pixel to the upper right of the pixel of interest. Encoding device.