JPH02274177A - Data compression method - Google Patents

Abstract

PURPOSE:To compress a halftone picture data efficiently by changing the data of a noticed block into a black level or white level when a prescribed pattern taking the noticed block as a white or black level isolated point takes place and applying the data compression. CONSTITUTION:After at two-dimension dot is converted into a linear density pattern, when a white or black isolated point exists as to a picture data reconstituted while being arranged adjacently to a prescribed location, smoothing processing is applied. Thus, the prescribed number of blocks including the noticed block in the data compression direction are referenced and when a prescribed pattern taking the noticed block as the isolated point takes place, the data of the noticed block is revised so that no isolated point takes place. Then a pattern having white and black levels caused alternately is reduced and the data compression efficiency is improved.

Description

[Detailed description of the invention] (Industrial application field) The present invention relates to a data compression method for halftone images.

(Conventional technology) When sending images by facsimile, etc., the sending side usually compresses and
A process called decompression is performed on the receiving side so that the number of bits actually sent is smaller than the number of bits in the image, thereby reducing communication time.

This compression/expansion follows certain rules (MH method, MR method, etc.)
This rule is based on the general manuscript (
It is designed to be more efficient when sending items (with characters, figures, etc. drawn on them).

That is, such a document has a very high probability of consecutive white or black (especially white) dots, and compression is performed by converting frequently occurring consecutive dots into data with a smaller number of bits. However, in such a compression method, there is a phenomenon in which the number of bits of data after compression conversion increases when the probability of occurrence is a few dots or less. This shortcoming usually does not pose a problem since overall compression efficiency is improved.

Incidentally, recently there has been an increasing demand for sending images containing halftones, such as photographs, using facsimiles, and some facsimiles have a photo mode.

Since only white or black can be expressed in bit units, halftone processing is performed on halftone images, and one pixel is treated as multiple bits (usually NXN bits), and the black and white within that pixel is The ratio expresses intermediate tones.

If you try to send a halftone image by facsimile, such image data has a lower probability of continuous white or black than a general text document, so conventional compression is inefficient and conversely There is also a possibility that the number of bits will increase.

In another application (Japanese Patent Application No. 1-44139)
Issue) discloses a data compression method that can efficiently compress halftone images. In this data compression method, as a preprocessing for data compression, an image is represented by a plurality of dots that have a two-dimensional spread corresponding to multiple levels of density, and a one-dimensional density pattern corresponding to this represented image is created. Let's decide. Furthermore, the reconstruction position of the image data is determined so that the one-dimensional density patterns are arranged adjacently. In this way, two-dimensional dots are converted into a one-dimensional density pattern at a predetermined position. The image data thus reconstructed is compressed. 1
Since the two-dimensional density patterns are arranged adjacent to each other, white or black dots tend to be continuous in the data compression direction, allowing efficient compression.

(Problems to be Solved by the Invention) The above data compression method tends to cause white or black dots to occur continuously in the data compression direction, making it possible to improve the compression efficiency of image data.

However, even with conversion to a one-dimensional density pattern, if white or black dots exist isolated in a halftone image, a pattern of alternating black and white tends to occur, and data compression efficiency cannot be improved in the vicinity. For example, in the MH method, white 1
The dot is 6 times the data, and the black one is! The dots are converted into three times as much data, increasing the number of data.

An object of the present invention is to provide a data compression method that can compress halftone image data more efficiently.

(Means for Solving the Problems) The data compression method according to the present invention expresses an image by a plurality of dots having a two-dimensional spread corresponding to a plurality of levels of density, and converts this expressed image into a Convert it to a dot pattern, arrange these one-dimensional dot patterns adjacently in a predetermined order, focus on one block in turn for each line in the data compression direction of the rearranged image data, and calculate a predetermined number of blocks before and after that block. If a predetermined pattern in which the block of interest is a white or black isolated point occurs, the data of the block of interest is changed to black or white, and then data compression is performed. It is characterized by

(Operation) After converting two-dimensional dots into a one-dimensional density pattern, smoothing processing is performed on the reconstructed image data arranged adjacent to a predetermined position if there is an isolated white or black point. For this reason, a predetermined number of blocks including the block of interest are referenced in the data compression direction, and if a predetermined pattern with the block of interest as an isolated point occurs, the data of the block of interest is changed so that the isolated point does not occur. do. This reduces the number of patterns in which black and white patterns occur alternately, improving the efficiency of data compression.

(Example) Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

The case of the density pattern method in which one pixel is composed of four 2×2 dots to express a halftone image will be explained. in this case,
Since a predetermined number of dots out of 4 dots are expressed in black corresponding to the pixel density (0 to 4) expressed in 5 levels, for example, as shown in Figure 1, 5N becomes blacker as the density increases.
Pixels can be expressed using similar density patterns (black dots are represented by hatched areas).

Therefore, the read halftone image is expressed as a set of density patterns, and if a part of it is extracted, for example, the second
It will look like the figure.

If the halftone image data shown in FIG. 2 is compressed as is using a conventional data compression method (for example, the MH method), the following disadvantages occur. For example, the top line and third line in the direction of lateral force in Figure 2 are almost the same! The black and white are reversed for each dot. In this case, 1 white bit is converted to 6 times as much data, and 1 black dot is converted to 3 times as much data, resulting in 4 or 5 times the amount of data, which is not suitable for the purpose of data compression. This will have the opposite effect.

In order to effectively compress halftone image data, in this embodiment, a two-dimensional 2×2 dot density pattern is converted to a 4×1 dot one-dimensional density pattern shown in FIG. 3, and then the density pattern shown in FIG. Perform smoothing as shown in .

In this conversion, as shown in Fig. 4, the dot positions of the 2 x 2 dot density pattern and the dot positions of the 4XI dot density pattern are successively converted to black dots in the 4 x 1 dot density pattern from the bottom. Correspond accordingly. Then, two one-dimensional patterns of 4×1 dots are arranged adjacent to each other in the data compression direction. Due to this adjacent arrangement, positions where black dots and white dots are likely to occur are adjacent to each other.

In image reconstruction using this one-dimensional pattern,
The 4Xl dot pattern is two 2×2 dot patterns, and the length of one side (4 dots) perpendicular to the data compression direction is the same. Therefore, in order to make the reconstructed image the same size as the original image, for example, as shown in FIG.
Extract the image data for the line in units of 2 x 2 dot patterns (that is, 1 pixel) in order from the left side to the top and bottom,
Convert it to a 4×1 dot pattern and arrange it to the right. Due to the adjacent arrangement of this one-dimensional pattern, positions where black dots or white dots are likely to occur are lined up adjacently.

FIG. 6 shows the image reconstruction results for the halftone image data of FIG. 2. Figure 7 shows the dot position (i) in Figure 2.
The correspondence with D is shown. The number ij is (
j+1)th data is shown.

It is clear that in the reconstructed image of FIG. 6, compared to the original image data of FIG. 2, white or black appears relatively continuously in the horizontal direction (data compression direction). Therefore, the conventional general data compression method (
MH method etc.), data compression can be effectively performed.

Next, smoothing is performed to eliminate isolated white or black points in the reconstructed image data to improve data compression efficiency. In this embodiment, the pattern-converted data is read out, the two blocks before and after the block of interest are referred to, and the data of the block of interest is changed only when the six patterns shown on the left side of FIG. 8 occur. . Do not change anything other than these patterns. These patterns are cases in which three or four blocks of the block of interest are either white or black.

In addition,! The two blocks at the beginning and the two blocks at the end of the line are smoothed by setting them to "0" (white) if there is no reference block.

FIG. 9 shows the result of smoothing the reconstructed image of FIG. 6 using the example shown in FIG.

For example, for a black isolated point on the first line, ooto
Since it corresponds to the pattern oo, it is changed to white data. In the case of a black isolated point at the beginning of the second line, “0010
1", so it is changed to white data. In the case of the next white isolated point, the pattern "01010" does not exist, so no change is made. In the case of the next black isolated point corresponds to the pattern 10100'', so it is changed to white data. Hereafter, smoothing is performed in the same way. Here, the two black isolated points corresponding to the toolot pattern are changed to white data.

Furthermore, the two white isolated points corresponding to the "0101 bi pattern" are changed to black data.The above smoothing eliminates the pattern in which black and white appear alternately.

Therefore, if a conventional general data compression method (such as the MH method) is used after smoothing, data compression can be performed more effectively.

Furthermore, when restoring the original image, the reverse conversion from 4×1 dot data to 2×2 dot data may be performed.

Note that instead of arranging two rows of pixels alternately in the one-dimensional pattern as shown in FIG. 5, the pixels in the [rows] may be arranged as they are. Although the reconstructed image has a different size from the original image, it can be restored to the original image on the receiving side.

As described above, by converting two-dimensional image data into a one-dimensional pattern, performing smoothing if necessary, and then compressing the data, compression is relatively efficient even when using conventional data compression methods. It is effective for high-speed transmission and storage in a small memory area.

Next, FIG. 1O shows the configuration of a facsimile machine that implements the data compression method described above. On the sending side of this facsimile machine, there is a reading section! reads the image, converts the density per pixel into a digital value, and sends it to the halftone processing section 2.
Then, this digital density value is converted into 2×2 dot halftone image data, and the 2D/1D converter 3 converts the 2×2 dot halftone image into a 4×1 dot image as explained above. The compression processing unit 4 compresses the converted data using a general data compression method (for example, the MH method) and transmits it. Further, the transmission data is stored in the storage unit 5 and transmitted.

゛On the receiving side, the communication data is stored in the storage unit 5 if necessary.
The decompression processing section 6 converts the compressed data into converted data, the one-dimensional/two-dimensional conversion section 7 converts the converted data into halftone data, and the output section 8 prints the image on paper.

The configuration of the facsimile machine shown in FIG. 1O differs from conventional facsimile machines in that it includes a two-dimensional/one-dimensional conversion section 3 and a one-dimensional/two-dimensional conversion section 7. Other parts, such as the compression processing unit 4 for data compression and decompression
, the decompression processing section 6, etc. are the same as those of the prior art.

Next, a conversion circuit controlled by a CPU (not shown) in the two-dimensional/one-dimensional conversion section 3 will be explained using the circuits shown in FIGS. 11 to 13.

To convert the halftone image shown in FIG. 2 into the reconstructed image shown in FIG. 9, the image data is written into the memory by the writing circuit shown in FIG. The data is read out and reconstructed as shown in FIG. In reality, the circuits in both figures can be configured together as one circuit. Finally, if the pattern corresponds to a predetermined pattern, smoothing is performed by the smoothing circuit shown in FIG. 13.

In the write circuit shown in FIG. 11, the address bus, data bus, and write signal (WR) of the CPU that controls the two-dimensional/one-dimensional converter 3 are connected to four memories 21, 22, 2.
3. Connected to the address terminal, data terminal, and shadow terminal of 24. Each memory 21.22.23.24 is
The data of the first, second, third, and fourth columns of the halftone image is stored. Therefore, the address bus signal is decoded by the decoder 25 to generate a chip select signal, and only one memory is selected by manually inputting the signal to the C8 terminal of each memory. In this way, four columns of signals are stored in memory 21.2 for each column.
They are written in the order of 2, 23, and 24.

The data in the memories 21, 22, 23, and 24 are read out and reconfigured by the follow-up circuit shown in FIG.

At this time, the data in each column must be rearranged after being read. As shown in Figure 7, the output data of the first line (use "line" for output data to avoid confusion with "column" for input data) is the output data of the first and third columns. This is a rearrangement of half of the human data. Therefore, in the circuit of FIG. 12, if the decoded signal of the decoder 25 that specifies the output line is the first line output signal for the first line, the chip select signal is sent to the memory 21.23 via the negative logic OR gate 31.33. simultaneously output to the CS terminal of As a result, data in the memory 2123 is simultaneously read out in response to read signals from the CPU. Furthermore, as shown in FIG.
．． ) and the odd-numbered bits of the third column (21, 23, .
) must be arranged alternately as shown in Figure 7. Therefore, the odd-numbered bits (data terminals !, 3, 5.7) of each of the memories 21 and 23 are arranged in the order of output data shown in FIG. 7 and connected to the input terminal of the 3-state gate 35. . The output terminal of this gate 35 is connected to a smoothing circuit (FIG. 13).

The gate 35 is also selected with the first line output signal of the decoder 25 as the gate control signal.

In this way, the data in the memories 21 and 23 are sequentially read out in accordance with the addresses and reconstructed into the output data of the first line.

The output of the second line is performed in the same manner. In this case, even-numbered bits (10, 12, . . .
・Even-numbered bits (3
0, 32,...) are combined and output. Therefore, in order to simultaneously select the memory 22.24 at the time of outputting the second line, the second line output signal of the decoder is outputted to the C8 terminal of the memory 22.24 via the OR gate 32.34. - On the other hand, the even-numbered bits (data terminals 0, 2, 4.6) of the memories 22 and 24 are combined to form the gate 36.
The smoothing circuit (FIG. 13) is connected to the smoothing circuit (FIG. 13).

Gate 3G is selected by the second line output signal of decoder 25.

The output of the third line is performed in the same manner. In this case, the odd-numbered bits (11', 13 . . .
) and the odd-numbered bits (31, 33, . . . ) of the fourth column data of the memory 24 are combined and output. Therefore, in order to simultaneously select the memories 22.24 at the time of outputting the third line, the third line output signal of the decoder is outputted to the C8 terminal of the memory 22.24 via the OR gate 32.34.

On the other hand, the odd-numbered bits (data terminal ratio 3, 5.7) of the memories 22 and 24 are combined and connected to a smoothing circuit (FIG. 13) via a gate 37.

Gate 37 is selected by the third line output signal of decoder 25.

The output of the fourth line is performed in the same manner. In this case, even-numbered bits (00, 02, . . .
) and even-numbered bits of the third column data of the memory 23 (
20, 22,...) are combined and output. Therefore, in order to simultaneously select the memory 2 23 at the time of outputting the 4th line, the 4th line output signal of the decoder is outputted to the C8 terminal of the memory 21.23 via the OR gate 31.33.

On the other hand, the even-numbered bits (data terminals 0, 2, 4.6) of the memories 21 and 23 are combined and connected to the smoothing circuit (FIG. 13) via the gate 38.

Gate 38 is selected by the fourth line output signal of decoder 25.

In the smoothing circuit shown in FIG. 13, the 8-bit data read out by the circuit shown in FIG. latched sequentially. Note that two pieces of white data (“0”) are sent before the first data of one line is sent and after the last data. The data latched in latches 52 to 56 is
It is sent to the address terminal of ROM57. This ROM57
Data is stored in such a manner that data smoothing is performed only in the case of the pattern shown in FIG. The data, which has been smoothed if necessary, is then sent to the S/P converter 58, converted into 8-bit parallel data, and then sent to the compression processing section 4.

In this embodiment, in the smoothing circuit (Fig. 13), data conversion by smoothing processing is unrelated to processing of other blocks, but data after data conversion is used for processing of other blocks. It is also possible.

Data successive output and data writing in the one-dimensional/two-dimensional converter 7 during demodulation are also possible by using the write circuit and the successive output circuit described above in reverse. (However, since the smoothing circuit is unnecessary, it is not used.) Next, a modified example will be described. When data is reconstructed and transmitted using the data compression method according to the present invention, the receiving side may use a normal facsimile machine that cannot convert the data to the original image. Even in such a case, it is desirable to be able to reproduce the image to some extent.

When the dot positions are converted as shown in FIGS. 3 and 4, the black dots tend to line up not only horizontally but also vertically. Therefore, it is better to use a method of converting into one-dimensional data that disperses black dots, as shown in FIG. 14, instead of a method that concentrates black dots in the vertical direction as shown in FIG. In the case of Fig. 14, when converting 2 x 2 dots into one-dimensional data of 4 x 1 dots corresponding to 5 levels of density θ ~ 4, black dots are distributed vertically as shown in Fig. 15. Configure.

When the data of the halftone image in FIG. 2 is reconstructed using the method shown in FIG. 14, the data is converted as shown in FIG. 16. Furthermore, if the smoothing method shown in FIG.
As shown in the figure, it is possible to reconstruct an image with fewer alternating black and white patterns. Therefore, if this reconstructed image is compressed and transmitted, even if it is decompressed on the receiving side and reproduced without using the one-dimensional/two-dimensional converter 7, the type of reproduction that concentrates the black dots in FIG. Compared to the constituent images, the horizontal white or black lines are less noticeable. Therefore, from a macroscopic perspective, it is possible to effectively recognize the original image (FIG. 2) to some extent.

For a specific image reconstruction circuit, it is only necessary to change the selection of the chip select signal and gate control signal from the output of the decoder 25 in FIGS. 1I and 12.

Above, we have explained the case of converting a 2x2 density pattern to a 4xl density pattern.
The density pattern of dots (N is an integer of 2 or more) is N'X
When converting to a one-dot density pattern, data reconstruction may be performed in the same manner.

In this example, the case of facsimile was explained, but
It can also be applied to compression when storing image data in memory, such as in a filing system.

Further, even if the configuration of one pixel is other than NXN (for example, N×M or non-rectangular), similar effects can be obtained by the same processing, and it is also possible to apply not only the density pattern method but also the dither method.

(Effect 7 of the invention) A halftone image is reconstructed by adjacent one-dimensional density patterns, and furthermore, if there is an isolated white or black point, the data is smoothed so that the isolated point disappears. Black dots become more likely to continue in the direction of data compression. Therefore, even if image data is compressed using a normal data compression method (such as the MH method), the image data can be compressed more efficiently.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an example of a 2×2 dot density pattern. FIG. 2 is a diagram of an example of an image after halftone processing. FIG. 3 is a diagram of an example of a one-dimensional density pattern for image reconstruction. FIG. 4 is a diagram showing the correspondence between a 2×2 density pattern and a 4×1a degree pattern. FIG. 5 is a diagram showing correspondence of data conversion positions in image reconstruction. FIG. 6 is a diagram of a reconstructed image of the image of FIG. 2. Figure 7 shows '? FIG. 6 is a diagram showing the correspondence of dots between the halftone image of FIG. JS2 and the halftone image of FIG. 6; FIG. 8 is a diagram of a smoothing pattern. FIG. 9 is a diagram of a reconstructed image that has been smoothed. FIG. 10 is a block diagram of a facsimile machine. FIG. 11 is a diagram of the write circuit of the two-dimensional/one-dimensional converter. FIG. 12 is a diagram of a continuation circuit of the two-dimensional/one-dimensional conversion section. FIG. 13 is a diagram of the smoothing circuit. FIG. 14 is a diagram of a modified example of a one-dimensional pattern for image reconstruction. FIG. 15 is a diagram showing the correspondence between the 2×2a degree pattern and the 4×1;i11 degree pattern. FIG. 16 is a diagram of a modified example of the reconstructed image. 3...2-dimensional/1-dimensional conversion section, 7...1-dimensional/2-dimensional conversion section. Patent source Minolta Camera Co., Ltd. Agent
Patent Attorneys: Suho et al., Figure 14, Figure 15, Figure 16, Figure 17, Teku-ni-Jiku-II (method) 1. Indication of the case, 1999 Special Application No. 096864 No. 2, Name of the invention Data compression method 3゜Relationship with the case of the person making the amendment

Claims (1)

[Claims] (1) An image is represented by a plurality of dots with two-dimensional spread corresponding to multiple levels of density, this represented image is converted into a one-dimensional dot pattern, and this one-dimensional dot pattern is After arranging the image data in a predetermined order, focus on one block for each line in the data compression direction of the rearranged image data, refer to the data of a predetermined number of blocks before and after that block, and change the block of interest to white or white. A data compression method characterized in that when a predetermined pattern of black isolated points occurs, the data of the block of interest is changed to black or white, and then data compression is performed.