JPH04270563A - Data compression system in picture processor - Google Patents

Abstract

PURPOSE:To attain efficient data compression by applying the proper compression method to each bit plane when a picture data has a data structure of plural bits. CONSTITUTION:The picture processor consists of a picture data reader 101, a picture data processing system 102, a picture data compression expansion processor 103, an input interface 104, a data compressor 105, a transmission line 106, a page buffer 107, a picture data storage device 108, a data expander 109, and a picture output device 111 or the like. A picture data is divided into plural bit planes in this processor and a different data compression system is applied to each bit plane. That is, for example, the run length encoding system is applied to a high-order bit plane, and the prediction coding system having plural prediction devices corresponding to the bit frequency pattern is applied to the low-order bit plane.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] The present invention relates to a compression method for image data having a data structure in which one pixel has a plurality of bits, and in particular to an image processing apparatus in which the data compression principle is different between an upper bit plane and a lower bit plane. This relates to a data compression method.

0002

[Prior Art] When a facsimile machine reads image information, compresses the data, and transmits the image, the run length of the white/black information is encoded because a typical document contains a lot of binary data such as text data. A run-length encoding method is used. This takes advantage of the fact that white data is continuous and black data is relatively short, and compresses the data by encoding the run length of continuous data (mainly white data).

[0003] Furthermore, as a method to further improve the efficiency of one-dimensional encoding such as run-length encoding, two-dimensional sequential encoding that utilizes correlation not only in the horizontal direction but also in the vertical direction has been carried out. In halftone images such as
A predictive encoding method is also used in which the value of a current pixel to be encoded is predicted based on the value of a pixel that has already been encoded, and the prediction error with this is encoded.

0004

[Problems to be Solved by the Invention] By the way, the run-length encoding method used in facsimiles is suitable for compressing text data, but it is not suitable for compressing multivalued bit structures such as halftone images. In this case, it is impossible for the value "0" to continue, so effective compression cannot be performed.

On the other hand, in compression methods such as predictive coding methods,
Although it can be an effective compression method for halftone images, it is disadvantageous in terms of cost performance when compressing text data. Further, when an input image contains a mixture of binary data such as text information and multi-value data such as picture image information, it has not been possible to deal with the situation using conventional individual compression methods.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to provide a data compression method that can efficiently perform data compression on image data whose input image has a data structure of multiple bits. do.

[0007]

[Means for Solving the Problems] The present invention provides an image processing device that compresses input image information, expands the compressed data, and outputs an image, in which one pixel has a data structure of multiple bits. It is characterized in that the data compression methods are different for the upper bit plane and the lower bit plane, for example, the upper bit plane is compressed by a run length encoding method, and the lower bit plane is compressed by a predictive encoding method.

[0008]

[Operation] In the case of image data in which one pixel has a data structure of multiple bits, the present invention divides the image data into multiple bit planes and performs a different data compression method on each bit plane. By applying a run-length encoding method to the bit plane and applying a predictive encoding method having a plurality of predictors corresponding to bit frequency patterns to the lower bit plane, data can be efficiently compressed. Especially in the case of gradation data such as plus 1 color or plus 2 color,
By applying the run-length encoding method to the color/monochrome information bit plane and the predictive encoding method to the gradation information bit plane, it is possible to perform optimal data compression for each bit plane. .

[0009]

[Embodiments] Hereinafter, embodiments will be explained with reference to the drawings. FIG. 1 is a diagram showing the configuration of an image processing apparatus to which the data compression method of the present invention is applied. In the figure, 100 is a document, 101 is an image data reading device (IIT), 102 is an image data processing system (IPS), 103 is an image data compression/expansion processing device, 104 is an input interface (I/F),
105 is a data compressor, 106 is a transmission line, 107 is a page buffer, 108 is an image data storage device, 109
110 is a data decompressor, and 110 is an output interface (I/F
), 111 is an image output device (IOT), 112 is a copy/print output, and 113 is a network interface (I/F).

Image data obtained by scanning the surface of the original 100 in the IIT 101 is subjected to image processing in the IPS 102 and is input to the image data compression/expansion processing device 103 through the input I/F 104. The image data from the input I/F 104 is compressed in the data compressor 105 as described later, and is stored in the page buffer 10 for each document.
7 or as is in the data decompressor 109
is transmitted to the I/F 110, decoded, and
It is output to OT111. Further, it is transmitted from the output I/F 110 to the image network through the network I/F 113. Conversely, data from the image network is taken in through the network I/F 113, data is compressed and expanded, and a copy is output from the IOT 111.
Alternatively, it can be printed out. Also, compressor 1
It is also possible to store the data compressed in step 05 in the image data storage device 108 and output the images all at once when necessary.

Data compressor 105, data decompressor 109
The configuration of is shown in FIG. data compressor 1
In 05, the input data is first processed in predictors 120 such that the output of each predictor 120 has a longer run length of "0" values than the input. Each predictor 120 has 15 predictors of four different types. The first type of predictor is the bit above (b
it・above) predictor. The bit-above predictor makes predictions by referring to the value of the pixel X at the same position in the line before the pixel P to be predicted.

The second type of predictor is the pre-bit predictor shown in FIG. 3(b). The pre-bit predictor makes predictions by referring to the value of the pixel X immediately before the pixel P to be predicted. The third type of predictor uses a font (5
element) predictor. The font predictor determines the value of the predicted pixel P based on statistical properties based on the states of the five reference pixels X1, X2, X3, X4, and X5. The fourth type of predictor is a halftone predictor as shown in FIG. 3(d). The halftone predictor performs prediction by using a pixel 5 to 16 bits before the predicted pixel P as a reference pixel on the same line.

[0013] Each predictor 120 refers to data 16 bits earlier or data in the previous line, and therefore has a memory therefor. These predictor selections initialize the predictor at the beginning of each line. The priority order at that time is as follows.

■5-element predictor ■Bits-above predictor ■Pre-bit predictor ■Shortest pattern length halftone predictor After initial setting, the selection of a predictor is performed so that the error is 0 depending on the initially set predictor. If there is one, it is continued to be used, and if an error occurs, the predictors are sequentially selected in the order of (1) to (2) above, and the optimal predictor is used. There is an error in the predictor you are currently using,
If the errors are the same even if predictors are selected sequentially, the currently used predictor is used. Thus predictor 120
Then, error data is output in 4-bit units for input 8-bit data.

FIG. 4 shows error data generated by the predictor 120, and when 8-bit data is sequentially input,
The predictor 120 selects the predictor using the method described above and outputs error data in units of 4 bits (nibbles). error is 0
In this case, zero nibble (0000) is output, and the bit where the error occurs becomes 1 and is output as, for example, (0100).

The encoder 121 encodes data based on this error data. In that case, a terminator code is assigned according to the type of error, and as shown in Figure 5, if there is one error bit, that is, the error data is 0.
In the case of 001, 0010, 0100, 1000, the terminator code is TA, excluding the 0 nibble,
When there are a plurality of error bits other than TA, the terminator code is TB.

Furthermore, in the case of terminator code TA, as shown in FIG. 6(a), the error data is tttt→a
Like a, i.e. 0001→00, 0010→0
It is encoded as follows: 1, 0100 → 10, 1000 → 11.

Furthermore, the terminator code TB is shown in FIG.
As shown in b), the error data is as tttt→bbbb, i.e. 0011→0100, 0101→0
101, 0110→0110, 0111→0111
,1001→1001, 1010→1010, 10
11 → 1011, 1100 → 1100, 1101 →
1101, 1110 → 1110, 1111 → 1111
It is encoded as

Using this terminator code, the run length (the length of the zero nibble) is coded on a case-by-case basis. For example, as shown in FIG. 7, if there is one error bit (terminator code TA), it is 10aa if the run length is 0, and 0r if the run length is 1 to 25.
rr rraa, if the run length is 26 to 63, 11rr rrrr tttt, if the run length is 64
-89 is expressed as 11ss ssss tttt. However, r, rrrrrr or rrrrrr is a value representing the run length in binary, and sssss
s is a value expressed in binary by subtracting 64 from the run length.

Furthermore, when the terminator code TB has a plurality of error bits, the run length is 0 or 1 as shown in FIG.
and 1 if the run length is 2 to 63.
1rr rrrr tttt. In addition, ttt
t is encoded as aa or bbbb shown in FIG. Large runs are too long to be encoded with the code described above, so they are encoded as multiples of 64 zero nibbles, 11rr rrrr 0000 as shown in Figure 9.
is encoded in In this case, rrrrrr is the binary value of the run length divided by 64, and the run length is 64.
Those above and below 4032 will be encoded.

An example of such encoding will be explained with reference to FIG. The errors in the prediction series created by the predictor are shown in Figure 1.
0(a), it is encoded as shown in FIG. 10(b). In FIG. 10(b), 0000 0001 in the first line indicates the normal mode in which the following code is predicted and then encoded, and is a code to distinguish it from cases where neither prediction nor encoding is performed. . In the second line, the zero nibble run length is 8 and the terminator is 0001 for the series of 0s from the first to the fourth byte, so the terminator code is TA, and it corresponds to the case where the run length is from 1 to 25. Therefore, 1 to 25 in Figure 7
The code is encoded with reference to the case of 0, then the run length 8 is expressed as a 5-bit binary number 01000, and the terminator is the aa value 11 corresponding to 0001 (see Figure 6).
a)), so 8 is 0010 0011.
encoded into a binary number of bits.

In the third line, the run length of the two nibbles from the second half of the fifth byte to the sixth byte is 1, and the terminator is 1111 (TB). This time the run length is 0
, 1 (Fig. 8), first, 011,
Next, bbbb representing the terminator is encoded as 0111 1111 by adding 1111 (FIG. 6(b)) corresponding to 1111 (FIG. 6(b)) and finally adding 1 representing the run length. The fourth line is included in the reappearing scan lines and represents the start of the next line.

The error data encoded by the encoder 121 in this way is transmitted through the transmission line, and is sent to the decoder 12.
3 to be decrypted. When the error data is decoded, the inverse predictor 124 converts the error data into image data based on the rules of the used predictor. In this case, even if data indicating which predictor is used is not separately transmitted, it is possible to determine which predictor is used by looking at the transmitted data. That is, the initially set predictor is determined, and if the error is 0, the initially set predictor is used, and if an error appears, the predictors are used in the order based on the above-described rules. Therefore, the used predictor is determined from the data, inverse prediction is performed based on the prediction rule of the used predictor, and the result is output as image data.

Next, using the data compression method as described above, the data compression method of the present invention will be described for the case where the input data to the image data compression/expansion processing device 103 is raster data and serial input. for example,
To explain the case where serial data in which one pixel data is 4 bits as shown in FIG. 11 is input, this serial input data is divided into four bit planes as shown in FIG. 12, and the bit string for each bit plane is Convert to Although not shown, such conversion can be easily realized by using a well-known 4-bit serial-to-parallel converter and forming a bit string for each 4-bit parallel output. Of course, one pixel may have 5 or more bits instead of 4 bits.

As shown in FIG. 13, the bit plane raster-scans the input image 130 and selects pixels P0, P1,
For P2, P3..., the plane of the least significant bit of each pixel is set to 131, and b00,
b10, b20, b30..., and similarly for the next upper bit plane 132, b01, b11,
b21, b31..., b02, b12, b22, b32 for the next higher bit plane 133
..., b03 for the most significant bit plane 134,
It is divided into four bit planes as b13, b23, b33...

In the present invention, image data is divided into bit planes as if there were four originals for one input original, and compression suitable for each bit plane is performed as shown in FIG. Apply the method. In other words, if it is binary data such as text data, run-length encoding can be used; if it is halftone image data, predictive encoding can be used, and a predictor suitable for the data can be used. good. Therefore, the compressor 105 of the image data compression/decompression processing device 103 in FIG. 1 is provided with predictors or compressors for the number of bits corresponding to each pixel.

Next, a case will be explained in which the image data has a color flag structure, for example, a case of plus one color (or plus two colors) having one color image for a white/black image. For example, one pixel data is bn0, bn1, bn2, bn as shown in FIG.
When set to 3, the lower 3 bits are density information G0, G1, G2.
, and if the most significant bit bn3 is the color flag, then
As shown in FIG. 15, the color flag bn3 is 1 or 0, and the lower three bits represent eight gradations from white to black. Further, in the case of red, it represents eight gradations from 0 to 111 in brightness.

By the way, as shown in FIG. 16, in a +1 color document, color images and black and white images are continuous, such as a color image area 140 in some parts and a black and white image area 150 in another part, and are also exclusive. Most of them exist locally and locally. Therefore, when 4-bit data is divided into 4 bit planes as mentioned above, red or black is generated locally in the most significant bit plane, with only a certain part being red and the rest being black and white. Binary processing is more suitable than half-pattern processing, and run-length encoding methods used in facsimile and the like are suitable.

On the other hand, for the bit planes bn2, bn1, and bn0, the respective bit frequency patterns are recognized, and a predictive compression method is used in which the above-mentioned easy-to-compress predictor is applied. In this way, in the case of plus one color or plus two colors, unlike a full color image, color information or black and white information exists continuously, exclusively and locally, so the run-length encoding method is optimal. In addition, even in the case of pseudo-halftone images and picture images, the data array of each bit plane becomes a certain periodic pattern or continuous 0 or 1 information when viewed locally, so a compression method suitable for each bit plane is used. By applying this, it becomes possible to perform efficient data compression.

[0030]

As described above, according to the present invention, when image data has a data structure of multiple bits, efficient data compression can be performed by applying a compression method suitable for each bit plane. becomes possible.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the configuration of an image processing device to which a data compression method of the present invention is applied.

FIG. 2 is a diagram showing the configuration of a data compression/expansion section.

FIG. 3 is a diagram for explaining a predictive encoding method.

FIG. 4 is a diagram for explaining error data.

FIG. 5 is a diagram showing a terminator code.

FIG. 6 is a diagram for explaining encoding of a terminator code.

FIG. 7 is a diagram for explaining run length encoding.

FIG. 8 is a diagram for explaining run length encoding.

FIG. 9 is a diagram for explaining run length encoding.

FIG. 10 is a diagram for explaining an example of encoding.

FIG. 11 is a diagram for explaining input serial data.

FIG. 12 is a diagram for explaining a data compression method.

FIG. 13 is a diagram for explaining the concept of a bit plane.

FIG. 14 is a diagram for explaining the data structure in the case of plus one color.

FIG. 15 is a diagram for explaining the data structure in the case of plus one color.

FIG. 16 is an explanatory diagram of a plus 1 color original.

[Explanation of symbols]

100... Original, 101... Image data reading device (IIT
), 102...Image signal processing system (IPS), 103
...Image data compression/expansion processing device, 104...Input interface (I/F), 105...Data compressor, 106...
Transmission line, 107... Page buffer, 108... Image data storage device, 109... Data decompressor, 110... Output interface (I/F), 111... Image output device (IO
T), 112...Copy/print output, 113...Network interface (I/F).

Claims (3)

[Claims] Claim 1: An image processing device that compresses input image information, expands the compressed data, and outputs an image, wherein one pixel has a data structure of multiple bits, and an upper bit plane and a lower bit plane are arranged. 1. A data compression method for an image processing device, characterized in that the data compression method is different in the following. 2. The data compression method in an image processing apparatus according to claim 1, wherein data is compressed using a run-length encoding method for upper bit planes and a predictive encoding method for lower bit planes. 3. The data compression method for an image processing apparatus according to claim 1, wherein the upper bit plane consists of color/monochrome information and the lower bit plane consists of gradation information.