JPS621372A - Multigradation picture signal encoder - Google Patents

Abstract

PURPOSE:To compress information of a multigradation picture signal by comparing a signal train obtained by binarizing individual picture elements of a display picture block with a standard binary signal train corresponding to an average density level of each block and encoding the state where the position of a disaccording picture element can be forecasted and the state where it cannot be forecasted. CONSTITUTION:A CPU 1 binarizes the multigradation picture signal by calculation of the average density level of the display picture block and comparison between this density level and a threshold value of a dither pattern ROM 7. The binary signal train and the standard pattern corresponding to the average density level are compared with each other to compare positions of disaccording bits and disaccording picture elements with contents of a forecasting ROM 9. The state where the number of disaccording picture element is 9, the state where the number of them is (n) and positions of them can be forecasted, and the state where the number of them is (n) and positions of them cannot be forecasted are defines as modes 1, 2, and 3 respectively. In the mode 1, the mode code and the density level code are outputted to a P/S converter 10; and in modes 2 and 3, the picture element number code, the number of disaccording picture elements, and the picture element position code are added to the output of the mode 1 and the result is outputted to the P/S converter 10.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a multi-gradation image signal encoding device that can efficiently transmit or store multi-gradation image signals.

[Technical background of the invention and its problems]

Conventionally, in facsimile machines, a document is scanned and the output signal is converted into a black and white binary signal, and then encoded based on various encoding methods to reduce the amount of information to be transmitted.
Efficient transmission methods have been implemented. The encoding method is MH (Modif-ied Huffman), which is recommended by CCITT (International Telegraph and Telephone Consultative Committee).
) encoding method and MR (Modified Read) encoding method are well known.

These encoding systems were developed to compress the signal band while keeping the amount of information equivalent by focusing on the probability distribution characteristics of the signal obtained by digitizing a black and white binary original.

However, in recent years, there has been a demand for processing documents and color images that include multiple gradations, such as photographs.

However, MH which was developed for 2 electric signal encoding
When a coding method or an MR coding method is applied to a multi-tone image, a problem arises in that sufficient signal compression efficiency cannot be obtained.

On the other hand, in display devices or recording devices that can only reproduce black and white binary levels, the dither method and the density pattern method are known as methods for reproducing halftones in a pseudo manner. The density pattern method treats NXN dots as one display screen block and associates this with one pixel constituting a halftone image, while the dither method treats a dot on the display device as one pixel of the halftone image. This corresponds to the pixel of . In the dither method, a multi-gradation image signal obtained by scanning an original image is compared with a predetermined threshold value corresponding to each pixel position, and white or black is determined depending on the magnitude of the signal. That is, the density level of each pixel obtained by scanning is expressed as xij (
However, ij indicates the pixel position. ) and a predetermined threshold value is C1j, the density level Xij is equal to the threshold value C1j
If it is larger, 1 is substituted for the output pixel Yij at that position, and if it is equal or smaller, 0 is substituted and displayed.

In this way, the image signal binarized by the dither method is usually represented by a signal sequence of 1 or O, but 1
Or, because the length of consecutive 0s is short, the conventional MH encoding method or MR encoding method has a problem in that the overall code length cannot be sufficiently compressed and an efficient code structure cannot be obtained. .

[Purpose of the invention]

The present invention has been made to solve the above-mentioned problems, and is capable of efficiently transmitting multi-tone image signals to shorten the transmission time, and to significantly reduce the amount of memory in an image file device. It is an object of the present invention to provide a coding device for a gradation image signal.

[Summary of the invention]

This invention uses a binary signal train obtained by comparing each pixel of each display image block divided into N pixels with a predetermined threshold value, and a predetermined standard binary signal for the average density level of each block. A multi-gradation image is created by comparing the columns and detecting the position of the mismatched pixel, separating the mismatched pixel into a state where the position is predictable and a state where it is unpredictable, and encoding each state appropriately. It compresses the information of the signal.

〔Effect of the invention〕

According to the present invention, a multi-gradation image signal having a huge amount of information can be compressed to a large extent, so that communication costs can be reduced by efficiently transmitting a multi-gradation image.

Furthermore, in image file devices, by reducing the amount of memory, the device can be made smaller and lower in price.

[Embodiments of the invention]

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

FIG. 1 shows a schematic configuration of an encoding device that realizes the halftone image encoding method according to the present invention by microcomputer control.
abbreviated as U) 1% ROM (Reaa only mem
2, RAM (Random access)
ss memory) 3, D/A converter (Digital
to person analog converter) 4, line memory 5, line memory control unit 6, dither pattern ROM 7, standard turn l'LOM 8, prediction RO
It is composed of an M9, P/8 (parallel-serial) converter 10. The CPU 1 controls the entire device and executes the encoding process by referring to the contents of the ROM 2. Here, a CPU 1 with an 8-bit data node is used. control software is stored in the RAM 3. The RAM 3 temporarily stores data necessary for the encoding process by the CPU 1.The D/A converter 4 uses an externally supplied image signal clock. , a multi-gradation image signal obtained by scanning each pixel with an external scanner synchronously is input, and the density information of each pixel is quantized as a 4-bit (16-level) image signal. The line memory 5 is composed of four line memories 5a to 5d, and includes a line control section 6 to be described later.
It writes the image signal digitized by the converter 4 under the control of
If it is 048 pixels, it has a capacity of 2048 pixels x 4 bits x 4 lines. The line memory control unit 6 performs control to write the image signal digitized by the converter 4 into the line memories 53 to 5d in synchronization with the image signal clock according to a write instruction from the CPU 1, and also when this writing is completed. It has a function of notifying the CPU 1 of the end of input at a certain point, for example, using an interrupt function.

The dither pattern ROM 7 is a table storing threshold levels for binarizing the image signal of 4 bits per pixel inputted from the line memory 5. As shown in FIG.
1 to 16 threshold values are stored in one block consisting of 4 thresholds. Here, by sequentially comparing the density level Xij of the image signal read out by the CPU 1, the dither pattern, and the threshold value C1j of the ROM 7 at the corresponding pixel positions, it is possible to obtain two temporary borrowed symbol sequences from the multi-tone image signal. can.

The standard pattern ROM 8 stores a pattern of 2 temporary borrowed symbols obtained when the density level of each pixel in a 4 x 4 block is almost uniform, and the density is as shown in Fig. 3(a) K. Seventeen standard binary signal patterns of levels θ to 16 are stored. Here, CPUIK compares the standard pattern Pij with the two temporary borrowed symbol sequences Yij obtained by dithering with respect to the average density level of each block, and detects and counts mismatched bit positions.

The prediction ROM 9 calculates the appearance probability Hij for each pixel position of a mismatched pixel when comparing the two temporary borrowed symbol sequences Yij obtained by dithering the image signal and the standard binary signal pattern Pij corresponding to the average density level of the block. are stored by average density level. Here, as shown in Figure 3(b),
For each average density level, the occurrence order Kij of mismatched pixels is written in the prediction ROM 9 according to the order of appearance probability of mismatched pixels for each pixel position.

The P/8 (parallel-serial) converter 10 is a CPU
It outputs the code encoded in 1 to the outside, and the CPU
The code output in units of 1 to 8 bits is converted into a serial signal in synchronization with the output clock.

The operation of this device will be explained below.

FIGS. 4(a), 4(b), and 4(c) are flowcharts of the control software for encoding processing stored in the ROM 2. FIG.

Initially, as shown in the 70-chart in Figure 4(a), KC
When a write instruction is issued from PU 1 to the line memory control unit, the multi-gradation image signal from the external scanner is transferred to the D/A
After being digitized by the converter 4, it is written into the line memories 58 to 5dK under the control of the line memory control section 6. When writing to the line memories 5a to 5d is completed,
Encoding processing is started upon input from the line memory control unit 6 at the end of writing.

Here, the CPU 1 first inputs a 4-bit image signal from the beginning of each line memory 5a to 5d. As a result, the density of each pixel of a 4.times.4 dot pixel matrix corresponding to a specific positional relationship is inputted, for example, as shown in FIG. 5(a). Next, the CPU 1 calculates the sum of the density levels Xij of each pixel in each block, and divides it by the number of pixels, 16 (rounding down to the nearest whole number), to calculate the average density level. The average density level for the first and second blocks is 4, the average density level for the third block is 5, and the average density level for the fourth block is 4.
The average density level for blocks is 3. This average density level is temporarily stored in R, AM 3.

Next, the CPUI determines the density level Xi of each pixel in block 1.
j is compared with the threshold value of the dither pattern ROM 7, and the multi-gradation image signal is binarized. Hereinafter, the process of binarizing this one block of multi-gradation image signals will be referred to as dithering. This dithering will be explained using the flowchart of FIG. 4(b). In the figure, the reference pixel address (
i, j) are sequentially dithered starting from (0, 0).

Dithering is performed by comparing the density level Xij of each pixel of one input block with the threshold value C1j of the dither pattern ROM7, and when Xij<C1jO, a binary signal yij=o,
When ij ≧Cij, the binary signal y:j is set to 1, and the pixel address ij is not incremented but 1j (4, 4
) ends dithering.

Next, the standard pattern Pij predetermined for each average density level shown in FIG. 5(b) K is compared with the two short signal sequences Yij obtained by dithering shown in FIG. Detects and counts bit positions.

The details of this operation will be explained using chart 70 in FIG. 4(C). In the figure, the reference pixel address (', j
) are compared sequentially starting from (0,0).

Then, the standard pattern Pij shown in FIG. 3(a) K corresponding to the average density level of each block and the binary signal sequence Xij obtained by dithering are taken out in order from R and AM 3, and P
When ij to Yij, add 1 to the mismatched pixel counter n,
The pixel address i at that time is stored in jtaAM3. FIG. 5(a) and FIG. 5(b) for the first block shown in FIG.

As shown in (C), as a result of the comparison of the standard pattern PijtZ tentative borrowed sign sequence Yij, all pixels match, so the count value of the mismatched pixel number counter becomes 0. In the case of the second block, the standard pattern Pij shown in FIG. 5(b) K was obtained by dithering the second block shown in FIG. C) Since the pixel position 5 indicated by the middle 0 mark is a mismatch, the count value of the mismatch pixel number counter becomes 1. In the case of the third block, see Figure 5 (
b) As a result of comparing the standard pattern pij shown in K with the 2 short signal sequence Yij obtained by dithering the third block shown in the same figure (C), it was found that the pixel position 5 indicated by the 0 mark in the same figure (C) and 15 do not match, so the count value of the mismatch pixel counter becomes 2. In the case of the fourth block, as a result of comparing the standard pattern P1j shown in FIG. 5(b) with the two short signal sequences Yij obtained by dithering the third block shown in FIG. Since pixel positions 4 and 8 indicated by O marks in C) do not match, the count value of the mismatch pixel counter becomes 2.

At the end of each block of the above-described mismatch pixel detection process, n mismatch pixel positions ij are stored in the RAM 3 along with the count value n of the mismatch pixel counter.

Next, the CPU 1 compares the occurrence position of the mismatched pixel obtained through the above-described processing with the contents of the prediction ROM 9. Prediction ROM
9, the probability of occurrence of mismatched pixels between the standard pattern Pij of each average density level and the dithered two short signal sequences Yij is falsified by storing the occurrence order of each pixel position based on statistical properties.

Details of this operation will be explained using chart 70 in FIG. 4(d).

In the process shown in FIG. 4, 1. Predicted RO
Separate the results of M9 and comparison into three states.

In the figure, first, the count value n of matching pixels is taken out from the AM3 and compared to see if n=Q. When n=0, the unmatched pixel is 01, that is, the standard pattern Pij and the two short signal sequences Yij obtained by dithering all match.
This state is set to mode=O and the process ends. In FIG. 5, the first block corresponds to this state.

In the case of n←O in FIG. 4(d), processing is performed to compare the occurrence position of the mismatched pixel with the contents of the prediction RAM 9.

First, the first mismatched pixel position ij is taken out from the RAM 3 as the comparison number counter mt-n. Next this ij
Then, Kij is extracted using the average density level of the block as a reference address. The contents of the prediction ROM 9 xiJ store the order of occurrence of mismatch for each pixel position for each average density level, so when the number of mismatched pixels is n, the contents of the prediction ROM 9 using the mismatch pixel position ij as a reference address is stored. x
If ij satisfies K(i, j)≦n, prediction is possible. Further, if K(i, j)>n, prediction is impossible. This comparison is performed sequentially for m=0, that is, mismatched pixel positions within one block, and when % x (i, j)>n is detected, the process is terminated at that point. Here, a block in which the occurrence positions of all n mismatched pixels can be predicted is set to mode = 1, and a block in which the occurrence positions of mismatched pixels are unpredictable is set to mode = 2, and the CPU
1 ends the process.

In FIG. 5(a), the second block is compared with the standard pattern Pij shown in FIG. 5(b) at an average density level of 4,
Although one pixel at pixel position 5 marked with a circle in FIG. 3(C) is different, pixel position 5 at level 4 of the contents of prediction ROM 9 shown in FIG. 3(b) has the first prediction order. Ashimoae=t
becomes. Similarly, the third block has an average density level of 5,
Comparing the standard pattern Pij shown in FIG. 3(b) with the two temporary borrowed symbol sequences Yij obtained by dithering, pixel positions 5 and 15 indicated by the 0 mark in FIG. 3(C) are different. pixel positions 5 and 1 of level 5 of the contents of prediction ROM 9 shown in b)
5 has a prediction rank of 2.1, and prediction is possible for both pixels, resulting in mode=1.

Next, the fourth block shown in FIG. 5(a) K has an average density level of 3, and when comparing the standard pattern Pij shown in FIG.
), pixel positions 4 and 8 indicated by the 0 mark are different, but the third
In the content of level 3 of the prediction ROM 9 in FIG. 9B, the predicted position is 3.1, and it is impossible to predict a mismatch at pixel position 5, resulting in mode=2.

As shown in the overall 70-chart in FIG. 4(a), as a result of comparing the mismatched pixel positions with the prediction ROM 9, each block was separated into three states: mode 1, 2*3, and therefore cptj l performs the following encoding process on these three states.

FIGS. 6(a) and 6(b) are a list of symbols used in this embodiment.

Figure (a) shows the mode code, and the mode in which the standard pattern PiJ and the two temporary borrowed code sequences Yij obtained by dithering all match.
= 1 is expressed as a pigeon, and a state in which mismatches of n pixels are detected but all n pixels are predictable is expressed as M.

A state in which mismatches of n pixels are detected and prediction of one or more of these pixels is impossible is expressed as M.

FIG. 6(b) shows density level codes L0 to Lll+ indicating the average density level within the block, pixel number codes N0 to N111 indicating the number of mismatched pixels within the block, and pixel position codes ^ which specify the position of the mismatched pixel. ~A1g are each represented by a 4-bit code.

FIG. 4(e) shows a detailed flowchart of encoding.

First, the CPU 1 determines whether the mode of the block to be encoded is 1, 2, or 3, and if mode=o, outputs the mode code ``pigeon'' and the density level code.

For example, in the case of the first block shown in FIG.
As shown in C), the mode code nMOs and the density level code "0100" are output from the CPU 1 to the P/8 converter 10.

Next, if mode=1, the mode code M, “ol”
outputs the density level code and pixel number code. For example, the fifth
In the case of the second block shown in FIG. 6(C), the mode code M1''O1'' and the density level code L4''
0100"K continues, and a pixel number code No'"oooo" indicating that the number of mismatched pixels is 1 is output from the CPU 1 to the P/S converter 10. In the case of the third block, as shown in FIG. (C) As shown in K, the number of pixels that do not match the mode code M-01'' and the density level code L-''0101 is 2.
The pixel number code N, "0001" indicating that the CPU
1 to the P/8 converter 10.

Also, in the case of mode~1, mode = 2, and in this case, CPU 1 has the mode code M! It outputs "10", a density level code, the number of mismatched pixels, mismatched pixels, and n types of pixel position codes that specify all pixel positions. For example, in the case of the fourth block shown in FIG. 5(a), as shown in FIG. 6(C), there is a mismatch between two pixels: the mode code M, "10", and the density level code, '0011'. The pixel number code N1'''0001'' indicating that the
00'' is output from the CPU 1 to the P/S converter 10.

The encoding of each block is manipulated in the same way. For example, if one line has 2048 pixels, it will be divided into 2 blocks, so the same process is repeated for one page in units of four lines to complete the encoding for one page.

By the way, since the density level of multi-gradation image signals such as photographs generally changes relatively slowly, for example, 4 x 4 pixels (0.5 mm x 0
．． Each pixel in each block of 5 m) often exhibits a nearly uniform density level, and the
~60% matches the standard pattern. On the other hand, most of the blocks that do not match the standard pattern only have mismatches of one to several pixels, and the positions of the mismatched pixels also tend to be concentrated at specific positions for each average density level. That is, if the density level within the block is approximately uniform, there is a strong tendency for pixels at the threshold value during dithering and the density levels before and after the threshold value to become inconsistent.

For example, if the average density level of a block is 4, the probability of mismatch between pixel positions with a threshold of 3.4.5 is high.

As for the predicted ranking of mismatched pixels for each average density level, the highest probability of mismatch can generally be determined by statistically examining the properties of various multi-tone images.

As described above, according to the present invention, encoding efficiency can be significantly improved compared to conventional MH encoding systems, MR encoding systems, and the like. Therefore, in facsimile machines and the like, it is possible to improve transmission efficiency and thereby reduce communication costs, and in image file machines and the like, it is possible to reduce the size and cost of the device by reducing the amount of memory. Note that the present invention can be modified in various ways without changing the spirit thereof.

For example, in the above embodiment, when encoding a block in which mismatched pixels occur, if even one pixel cannot be predicted for n mismatched pixels, 9 pixel positions including the predictable pixels are encoded. However, in this case, only the number of mismatched pixels and unpredictable pixel positions may be encoded. In this way, the compression ratio can be further improved.

Further, in the above embodiment, the average density level, the number of mismatched pixels, and the position of the mismatched pixels are encoded for a block in which mismatched pixels occur, but the number of mismatched pixels does not necessarily need to be encoded. For example, it is also possible to encode using three types of codes indicating the average density level, the mismatched pixel position, and the end of the pixel position.

In addition, in the above embodiment, a multi-gradation image signal is transmitted per pixel (#)
Although the case of quantizing with 4 pits and dividing into 4×4 blocks has been described, the present invention is not concerned with the number of quantization bits or the number of pixels in a block.

Further, the arrangement of threshold values for dithering and the assignment of codes are not limited to the above embodiments.

Further, the contents of the prediction ROM are not limited to the above embodiments, and a plurality of prediction ROMs can be prepared and switched according to a document.

Furthermore, although the above embodiment describes a case of a black-and-white multi-gradation image, the present invention is not limited to this, and the present invention is not limited to this.
The present invention can be applied to each of the following.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, and an explanatory diagram showing encoding obtained by the example. ]...Microcomputer 2...ROM 3...RAM4...
D/A converter 5, 5a to 5d...Line memory 6...Line memory control unit 7...Dither pattern ROM8...Standard pattern OM 9...
Prediction FLOM10...P converter (≦shin <b) (α)
Figure 4 (,C) (tL) 4th
Figure (e) Figure 4 Figure 5

Claims (3)

[Claims] (1) A first means of inputting a multi-gradation image signal and dividing it into blocks consisting of N pixels, and comparing each pixel of each block with a predetermined threshold value.
a second means for converting into a value signal sequence; a third means for obtaining an average density level of each block; a fourth means for obtaining a predetermined standard binary signal sequence corresponding to the average density level; a fifth means for detecting mismatched pixels by comparing the binary signal sequence obtained by the second means with a standard binary signal sequence corresponding to the average density level of the block at this time; a sixth means for predicting the pixel position within the block of the mismatched pixel detected by the means; and the mismatched pixel detected by the fifth means, and according to the prediction result of the mismatched pixel obtained by the sixth means. and seventh means for appropriately encoding pixels of each block to compress information. (2) The fifth means separates the block into a first state in which all pixels match and a second state in which at least one pixel does not match, and the block in the first state encodes an average density level; In state 2, a block in which all mismatched pixels could be predicted by the sixth means encodes the average density level and the number of mismatched pixels, and a block in which no mismatched pixels could be predicted encodes the average density level and the mismatched pixel position. A multi-gradation image signal encoding device according to claim 1, characterized in that: (3) The fifth means separates the block into a first state in which all pixels match and a second state in which at least one pixel does not match, and the block in the first state encodes an average density level; The block in the second state encodes the average density level, the number of mismatched pixels, and the pixel position of the unpredicted pixel according to the prediction result of the mismatched pixels by the sixth means. 2. The multi-gradation image signal encoding device according to item 1.