JP3323615B2 - Image data compression apparatus and method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for compressing image data.

[0002]

2. Description of the Related Art In recent years, images created on a computer by DTP or the like are required to have higher image quality, and colorization and multi-gradation are being promoted. The information amount in this type of image is, for example, A4 size 400 dpi 25
In the case of 6 gradations and 3 colors, it is about 46 Mbytes.
Image data is written in a page description language (Page Discri
If handled as ption Language) code information such as the amount of information that a few. However, there is a problem that it takes time to develop the code information into the image data and that the original code information cannot be reproduced from the developed image.

As a general compression method of a color multi-valued image, JPEG (Joint Photographic) is used.
ADCT (A) recommended by the Expendable Group
Adaptive Discrete Cosine T
(transform) method. ADCT
The compression method will be described below.

FIG. 13 is a block diagram showing a functional configuration of the ADCT compression apparatus. Reference numeral 3101 denotes a color space conversion unit, which is a color space (NTSC-RGB) using signals (red (R), green (G), and blue (B)) for each color in the NTSC system.
Is converted into a color space (YCrCb) represented by a luminance signal Y and two color difference signals Cr and Cb. Reference numeral 3102 denotes a sub-sampling unit, which reduces the color difference data by using the characteristic that the human eye is sensitive to luminance and insensitive to color difference. Specifically, the sub-sampling unit 3102 takes the average value of two adjacent color difference data, and reduces the amount of the color difference data to half. Reference numeral 3103 denotes a DCT conversion unit which divides the image data input via the sub-sampling unit 3102 into 8 × 8 blocks that are adjacent in the horizontal and vertical directions, performs DCT conversion, and converts the data into a frequency space. Reference numeral 3104 denotes a quantization unit, and each of the 64 DCTs
The coefficients are divided by different quantization values of the step size. 3105
Is a Huffman encoding unit, and 64 quantized DCs
The T coefficient is divided into one DC coefficient and 63 AC coefficients, and each coefficient is encoded according to a Huffman table recommended by JPEG. The encoded data is stored in a memory with a header such as quantization table data and Huffman table data attached thereto, or transmitted to another device or the like.

FIG. 14 is a block diagram showing a functional configuration of the ADCT decompression device. Reference numeral 3205 denotes a Huffman demultiplexing unit that demultiplexes the input coded data to generate quantized data. An inverse quantization unit 3204 converts the quantized data generated by the Huffman decompression unit 3205 into CDT coefficient data. This is based on the quantization table data used for quantization in the quantization unit
It is obtained by multiplying the four coefficients by the quantization value. 3
An inverse DCT transform unit 203 converts the DCT coefficient data obtained by the inverse quantization unit 3204 into real image data by inverse DCT.
Convert. An interpolation unit 3202 interpolates the data of Cr and Cb missing by the sub-sampling unit 3102 at the time of data compression by a simple repetition method.
A color space conversion unit 3201 converts YCrCb data into NTSC-RGB data or color space data of the device.

Next, the flow of processing of the present data compression and decompression device will be described with reference to actual data.

FIG. 15 shows a part of image data of a color multi-valued image created by a computer. FIG.
Is NTSC-RGB data, which is one portion (16 pixels × 16 pixels) of image data in which a character portion of a color multi-valued image is developed. 3301
Is red (R) data, 3302 is green (G) data, 330
3 indicates blue (B) data. The data of each pixel is data having a range of 8 bits (0 to 255), and (R, G,
B) = (225, 225, 225) is a part of an image in which blue characters (R, G, B) = (30, 30, 225) are written on a slightly darker white background.

[0008] NTSC performed by the color space conversion unit 3101
-Conversion from RGB to YCrCb is performed by the following equation.

Y = 0.299 × R + 0.587 × G + 0.114 × B Cr = 0.713 (RY) Cb = 0.564 (BY)

Further, according to CCIR's recommendation, YCrCb
The following data is rounded to allow overshoot and undershoot at the time of calculation in the color space.

Y = 219.0 × Y + 16.5 Cr = 224.0 × Cr + 12.5 Cb = 224.0 × Cb + 128.5

Each of the YCrCb data obtained by the above equations is converted by the sub-sampling unit 3102 into the color difference data Cr.
And Cb are sub-sampled. As the sub-sampling method, simple thinning, MAX data selection, M
Although there is IN data selection, here, the average value method is used. That is, this method is a method in which an average value of data of two connected pixels is taken as one data. FIG. 16 shows data obtained through the above color space conversion unit 3101 and subsampling unit 3102. In FIG.
01 is Y data (luminance data), 3402 is Cr data (color difference data), 3403 is Cb data (color difference data)
It is. Each data of Cr and Cb is stored in the sub-sampling unit 3.
It can be seen that the value is reduced to 処理 by the processing of 102.

Next, the Y data 3401 shown in FIG.
Cr data 3402 and Cb data 3403 are input to DCT transform section 3103. First, the DCT transform unit 3103 divides each data into blocks each of which is 8 × 8 data that is horizontally and vertically. As a result of the block division, the Y data 3401 is changed from 3401a to 3
It is divided into four blocks 401d. Similarly,
The Cr data 3402 is divided into two blocks 3402a and 3402b, and the Cb data 3403 is divided into 3403
a, 3403b are divided into two blocks. Then, DCT transform is performed on each of these eight blocks.

FIG. 17 shows that the eight blocks shown in FIG.
The state after the T conversion is shown. Reference numeral 3501 denotes data after the DCT conversion of the Y data 3401. Four blocks 3401
3501a to 3501d for a to 3401d respectively
Corresponds to the block. Similarly, reference numeral 3502 denotes data after DCT conversion of the Cr data 3402, and reference numeral 3503 denotes data after DCT conversion of the Cb data 3403. DC
The 64 coefficients of each block after the T conversion are composed of one DC component (upper left corner) and 63 AC components.

Next, the DCT transformed data 350 shown in FIG.
1 to 3503 are quantized by a quantization unit 3104.

The quantization table used for quantization is J
Use the quantization table recommended by PEG. FIG. 22 shows a quantization table. 4001 is for the Y component, 4002
Is for Cr and Cb components. FIG. 18 shows the data after the quantization processing. 3601 is the quantized data of Y data, 3602 is the quantized data of Cr data, 3603
Is the quantized data of the Cb data.

In the Huffman encoding unit 3105, each of the quantized data 3601 to 3603 is divided into a DC component and an AC component. For the DC component, a histogram of the difference from the DC component of the previous block is obtained, an optimal Huffman encoding table is created, and encoding is performed according to the table. A
The C component is rearranged in a zigzag order as shown in FIG. Then, for the AC component, a histogram of a combination of a run length of 0 and a value X until a value X other than 0 enters, and an optimum Huffman table is created. Encoding is performed according to the generated optimal Huffman table.

At this point, the data of 16 × 16 pixels in NTSC-RGB becomes 795 bits. Since the original image has 8 bits per pixel, 16 × 16 × 3 colors = 7
86 bytes = 6144 bits. Therefore, 1 /
7.7 compression has been achieved. Actually, an image size, a quantization table, an encoding table, and the like are attached to the encoded data, and are stored in a memory or transmitted.

Next, the processing at the time of data expansion will be described.

The image data processed by the ADCT compression method is input to a Huffman decomposing unit 3205 and decomposed. The inverse quantization unit 3204 includes a Huffman combining unit 3205
The multiplication data is multiplied by the coefficient of the quantization table shown in FIG. With the above processing, data as shown in FIG. 19 is obtained. FIG. 19 shows data obtained by executing the Huffman compounding process and the inverse quantization process on the data obtained by the above-mentioned ADCT compression method. When the data in FIG. 19 is compared with the data before compression in FIG.

Next, the data of FIG.
At 202, it is converted into YCrCb data. The YCrCb
The data is shown in FIG. Then, the YCrCb data in FIG.
It is converted to GB data. FIG. 21 shows data finally obtained by the data decompression process.

[0022]

As described above,
The ADCT compression method is an irreversible compression method involving data loss at the time of quantization in the sub-sampling unit 3102 and the quantization unit 3104. Therefore, the decompressed data is different from the data before compression. FIG. 15 shows NTSC-RGB data before the above-described compression processing.
This is apparent from a comparison of FIG. 21 with NTSC-RGB data obtained as a result of performing the compression / decompression processing on the data in FIG. And, this means that the image quality degradation occurs.

An image (computer-generated image) created on a computer, such as DCP data, has the advantage of clear outlines and noise-free coloring with a single color of one figure (or one character). However, when the image data is processed by the ADCT compression method, only the outline of the figure is blurred and a pseudo edge (called mosquito noise) is generated.
And the change in color due to quantization occurs, and the advantages of computer-generated images are not exhibited. In particular, since the above-described compression method is an 8 × 8 block process, the color greatly changes in the boundary area. When the compression ratio is increased, the AC component is finally lost, block distortion occurs, and an image having a resolution of 1/8 is obtained.

[0024] For this reason, it is conceivable to compress the image by quantization so that the deterioration of the image quality is not visible. However, since the image is originally a computer-generated image characterized by a large amount of high-frequency data, the compression ratio is low. Not practical.
Therefore, it is necessary to compress a computer-created image such as an image created by DTP by a compression method capable of obtaining a reversible and high compression ratio.

The present invention has been made in view of the above problems, and has as its object to provide an image data compression method and apparatus for compressing a multi-valued image created by a computer or the like at a high compression ratio without deterioration. I do. Specifically, when differentially encoding a multi-valued image on a pixel-by-pixel basis, the value of the encoding target pixel "is likely to be the same value as the previous pixel"; and
An object of the present invention is to perform efficient differential encoding focusing on the fact that the value of a pixel to be encoded is “there is a tendency to return from the pixel value that appeared two times before to the pixel value that appeared immediately before”.

[0026]

According to the present invention, there is provided an image data compression method comprising the steps of: inputting pixel data d [i] in units of one pixel; ] And preceding pixel data d [i−
1], a difference output step of outputting a difference diff with respect to
A first determining step of determining whether the difference diff is 0, a second determining step of determining whether the difference diff is x [k], and a first and a second determining steps. Based on the match / mismatch in the judgment, the d
And a code output step of outputting a code indicating [i]. When the difference diff is not 0 in the first determination step, x [k + 1] = − diff is updated, and the next pixel data d [ i + 1].

[0027]

FIG. 24A is a diagram showing the characteristics of an image created by a computer. The vertical axis of FIG. 24A indicates the value of the data, and the horizontal axis is the spatial axis. Computer-generated images tend to have the same data value as the first feature, and return to the original data even if the data changes as the second feature.

Therefore, as the height of the appearance probability, 1) the same data as the previous pixel (the difference is 0) 2) The vector opposite to the vector at the time of the pixel value change is likely to occur.

That is, if the difference has a change of plus 2, a change of minus 2 tends to occur. This is shown in FIG.
This will be described with reference to FIG.

If a vector of plus 64 is generated at 1101, the vector that is likely to occur next to the difference <0> is <−64.
>. In the example of FIG. 24B, a minus 64 vector is actually generated in the next 1102. The next most likely vector from the minus 64 vector of 1102 is <+64>. At step 1103, a plus 64 vector is generated. In the next 1104, a plus 191 vector was generated. At this time, the event that is likely to occur next to <0> is <-191>, and the next is <-255> calculated by -191-64.

The events that are likely to occur from the above characteristics can be expressed by the following formulas using a calculation formula. The spatial change is i,
The pixel value at that time is represented by d [i] (8 bi of 0 to 255).
t), k is incremented by the same color as a lump (see FIG. 24), and the next most likely vector after <0> is x
Assuming that the next most likely vector after [k], <0> and x [k] is represented by y [k], diff = d [i] −d
Assuming that [i-1], when diff is 0, x [k], y
[K] has no change. When diff is other than 0, x [k],
y [k] is updated as follows.

X [k + 1] =-diff, y [k + 1]
= Diff + A [k] A [k] is y when the generation vector matches x [k]
[K], otherwise x [k].

For i [0], x [0] and y [0], predetermined values are substituted as initial values. For example, i [0] = 0,
It is substituted as x [0] = 255 and y [0] = 100.

The data can be compressed by assigning a short code to an easily occurring event.

For example, the code when the difference is 0 is “0”, and the code when the difference matches x [k] is “1”.
0 ”and the code when the difference matches y [k] is“ 11 ”.
0, and data indicating a value is attached to a code “111” indicating that when none of them match, the data indicating this value may be a difference or color data. However, in this example, color data is attached, and an image whose value changes as shown in FIG.

Since d [1] -d [0] is 0, <0>
Is output.

Since d [2] -d [0] is also 0, <0>
Is output.

D [3] -d [2] is 64 and x
[0] = 10, y [0] = 5, so <111+
64>, that is, <1111000000>.

This change causes x [1] to be -64 (-d
iff) is substituted and 191 (−diff +
x [0]).

Since d [4] -d [3] is 0, <0>
Is output.

Since d [5] -d [4] is also 0, <0>
Is output.

D [6] -d [5] is -64 and x
Since it matches [1], <10> is output.

By this change, x [2] is changed to +64 (−d
iff) is substituted and 255 (−diff +
y [1]) is substituted.

Since d [7] -d [6] is 0, <0>
Is output.

D [8] -d [7] is 64 and x [2]
And <10> is output.

This change causes x [3] to be -64 (-d
iff) is substituted and 191 (−diff +
y [2]) is substituted.

Since d [9] -d [8] is 0, <0>
Is output.

D [10] -d [9] is +191 and y
Since it matches [3], <110> is output.

This change causes x [4] to be -191 (-
diff) is substituted and −255 (−dif) is substituted for y [4].
f + x [3]).

Since d [11] -d [10] is 0, <
0> is output.

Since d [12] -d [11] is also 0, <
0> is output.

Since d [13] -d [12] is also 0, <
0> is output.

Since d [14] -d [13] is also 0, <
0> is output.

Since d [15] -d [14] is also 0, <
0> is output.

Since d [16] -d [15] is also 0, <
0> is output.

Since d [17] -d [16] is -191 and coincides with x [4], <10> is output.

By this change, x [5] is added to +191 (−
diff) is substituted and y [5] is −64 (−diff
+ Y [4]).

By such processing, x [k], y
The value of [k] changes as shown in FIG. Also, the code data is 59 bits as shown in FIG. 24 (d). In this example, since the original data is 8 bits, 36 pixels × 8 =
What needed 288 bits is about 1/5 of 59 bits
Can be compressed. It can be applied to any bit image.

The situation when the data of FIG. 24D is expanded is shown below.

In the decompression process, d [0] = 0 and x [0]
= 255 and y [0] = 100, the same initial values as in the compression processing.

First, since the first code is <0>, d
[1] = d [0] = 0 can be derived.

Since the next code is also <0>, d [2] = 0
Can lead.

The next code is <111>, which is x
[0] and y [0] indicate that different data has been generated, so the next 8 bits (8 bits in this case, but 24 bits if the original data is 24 bits) are read out, and a numerical value of 64 is obtained. Substituting, d [3] = 6
It becomes 4. At this time, x [1] has-(d [3] -d
[2]) = − 64 is substituted, and y [1] is −64 + x
[0] = 191 is substituted. In this example, since color data is attached, d [3] = (read-out numerical value), x [1]
= − ((Read numerical value) −d [2]), but when a difference is attached, d [3] = d [2] + (read numerical value), and x [1] = (Read numerical value).

Since the next code is <0>, d [4] = 6
4 leads.

Since the next code is also <0>, d [5] = 6
4 leads.

The next code is <10>, which indicates that the difference is equal to x [1], so that d [6] = d
[5] + x [1] = 0 can be derived. At this time, x [2] is-
(D [6] −d [5]) = 64 is assigned, and 64 + y [1] = 255 is assigned to y [2].

Since the next code is <0>, d [7] = 0
Can lead.

The next code is <10>, which indicates that the difference is equal to x [2], so that d [8] = d
[7] + x [2] = 64 is derived. At this time, x [3] is-
(D [8] −d [7]) = − 64 is substituted, and y [3]
Is assigned to −64 + y [2] = 191.

Since the next code is <0>, d [9] = 6
4 leads.

The next code is <110>, which indicates that the difference is equal to y [3], so that d [10] =
d [9] + y [3] = 255 is derived. At this time, x [4]
Is substituted with-(d [10] -d [9]) = 191, and y
-191 + x [3] =-255 is substituted for [4].

With the above processing, the image data of FIG. 24A can be restored from the code of FIG. 24D without data loss.

Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings.

[Embodiment 1] A data compression apparatus according to Embodiment 1 will be described below.

FIG. 1 is a block diagram showing a functional configuration of the data compression device according to the first embodiment.

Reference numeral 101 denotes a unit for rearranging the raster image data input by the pixel data rearranging unit, and outputting the raster image data one pixel at a time in synchronization with the clock. Here, there is no problem even if there is no processing (the data may be output as a raster as it is), or the scan sequence may be changed and output as shown in FIG.
For example, in order to rearrange the pixel data as shown in FIG. 26A, the pixel data rearranging section is constituted by a memory, holds pixel data for eight lines in the vertical direction, and sequentially outputs data rearranged by address control. A compression unit 102 performs compression processing on the pixel data output from the pixel data rearranging unit 101 and outputs a code. 103, a data pack unit;
And outputs the packed data according to the data output from the memory interface or the communication interface. 104 stores or transmits data packed in a memory or a communication interface unit.

The compression section 102 will be described with reference to FIG. Reference numeral 201 denotes a 24-bit flip-flop that stores a pixel data value input in synchronization with a clock. An initial value is required. Here, as an example, white (255, 255, 25
5) is stored as an initial value.

Reference numeral 202 denotes pixel data d input by the subtractor
[I] and pixel data d [i−1] one pixel before, which is an output from the flip-flop unit 201, are input, and d [i] −
d [i-1] is calculated and 25-bit difference data dif
Output f.

An OR circuit 203 inputs the lower 24 bits of the diff output from the subtracter 202, performs an OR operation, and outputs 1-bit data code0. code
0 is 0 when diff is 0, and 1 when diff is other than 0.

Reference numeral 204 denotes an adder, which inputs the NOT of the diff output from the subtractor 202 and adds 1 to -diff.
Is output.

Reference numeral 205 denotes a 25-bit flip-flop to which -diff which is the output of the adder 204 is input, and NOT
When code0, which is the output of the circuit 203, is 1, the signal "-diff" is stored.
An initial value is necessary. Here, as an example, the difference (−255, −255, −255) when white changes to black is stored as the initial value. Reference numeral 206 denotes a comparator which receives the diff output from the subtracter 202 and the X output from the flip-flop 205 and outputs 0 if they match and 1 if they do not match.
It outputs data code1 of it.

A selector 207 is a flip-flop 20
Under the control of code1, which is input from the output X of 5 and Y described later and output from the comparator 206, X is selected when code is 1 and Y is selected when code1 is 0 and output.

Reference numeral 208 denotes an adder which receives the output of -diff of the adder 204 and the output of the selector 207, and
The addition result of t is output.

Reference numeral 209 denotes a 25-bit flip-flop to which the output of the adder 208 is input, and that code 0, which is the output of the NOT circuit 203, is used as a high enable signal.
When 0 is 1, data is stored. An initial value is required. Here, as an example, the difference (0, -2) when the color changes from white to red.
55, -255) are stored as initial values.

Reference numeral 201 denotes a comparison unit, which is a 1-bit value of 0 if the output diff of the subtractor 202 and Y the output of the flip-flop 209 match, and 1 if they do not match.
Is output.

Reference numeral 211 denotes a code allocating unit for code 0, c
mode1 and code2 are input, assigned codes, and output.

In this compression apparatus, there are four cases: 1) the difference diff is 0, 2) the difference diff matches X, 3) the difference diff matches Y, 4) none of 1), 2) and 3). For example, <0> for 1), <10> for 2), <11> for 3
If a code to be <111> is fixedly given to 0> and 4), the code allocating unit 211 is not necessary, and <code0> is used in 1) and <code1, cod is used in 2).
de1> in case of 3) and 4) <code0, code
1, code2> may be output. The code assignment unit 211 is necessary when it is desired to change the code according to the situation.

FIG. 6 is a flowchart showing the operation of the compression apparatus.

FIG. 3 shows an example of a decompression device for decompressing data compressed by the present compression device.

A memory or communication interface 301 outputs read data or received data. Reference numeral 302 denotes a data unpacking unit which is configured by a multiplexer or the like and unpacks and outputs the received compressed data. Reference numeral 303 denotes input of the data unpacked by the expansion unit, expansion to pixel data, and output. 304, the data expanded by the pixel data rearranging unit is output in the horizontal and vertical directions in the image space. This is not necessary if there is no pixel data rearranging unit 101 during the compression process. For example, FIG.
It is necessary to return to the original raster order when rearranged as shown in FIG.

The extension section 303 will be described with reference to FIG. Reference numeral 401 denotes a code return unit that is necessary when code conversion is performed by the code allocation unit during compression processing, but other than that is not necessary. It outputs 24-bit color data new attached to code0, code1, and code2.

Reference numeral 402 denotes a 24-bit flip-flop which stores pixel data d [i]. The same initial value as the initial value stored in 201 during the compression processing is required (255, 255,
255) is stored.

Reference numeral 403 denotes a subtractor for pixel data d [i-1].
-Dif of 25 bits which is a value obtained by subtracting d [i] from the
Output f.

Reference numeral 404 denotes a 25-bit flip-flop to which -diff output from the subtractor 403 is input and code
When "1" is 1 and "code1" is "1", "-diff" is stored. The same initial value as the initial value stored in 205 during the compression processing is required (−255, −255, −255).
255) is stored.

Reference numeral 405 denotes a selector for inputting X and Y, which will be described later, and controlling code1. When code1 is 1, X is set to c.
When mode1 is 0, Y is selected and output.

Reference numeral 406 denotes an adder which receives the output of the selector 405 and the output of the subtractor 403, ie, -diff, adds and outputs the result.

Reference numeral 407 denotes a 25-bit flip-flop which inputs the output of the adder 406, stores data when code1 is 1 with code1 as a high enable signal. The same initial value as the initial value stored in 209 at the time of the compression processing is required, and (0, -255, -255) is stored.

A selector 408 is a flip-flop 40.
X which is the output of 4 and Y which is the output of the flip-flop 407 are input, and under the control of code1, X is selected when code1 is 0, and Y is selected when code1 is 1, and output.

An adder 409 is a flip-flop 402.
, And the output of the selector 408 is input and added to output 24-bit data A.

Reference numeral 410 denotes an AND circuit for code 0, code
e1 and code2 are input and 1 is output only when all are 1.

A selector 411 receives the color data new, the output d [i-1] of the flip-flop 402 and the output A of the adder 409, and inputs code0 and an AND circuit 410.
If (code0, b) is used as the control of output b, (0, X), d [i-1], (1, 0) A, (1, 1) new, and color Data d
[I] is output.

FIG. 7 is a flowchart showing the operation of the present decompression device.

[Embodiment 2] In the above-described embodiment, two vectors x and y are given as signs with a high appearance probability. However, the present invention is not limited to this. The formula for calculating the vector is as follows. In the above-described embodiment, the vector represented by x [k] and y [k] is represented as vector [m] [k] in the present embodiment. m = 0 to n
When m = 0, it corresponds to x in the above embodiment, and m = 1
At this time, it corresponds to y in the above embodiment. k is a value that increments the same color as one lump (see FIG. 24A). The color change is diff = d [i] -d [i-1] (i
Is the spatial change of the pixel, d [i] is the pixel value at the position of i), and the vector vect [m] [k] is diff1 =
When it is 0 (when a color change occurs), it is updated as follows.

Vector [0] [k + 1] = − diff; vector [1] [k + 1] = − diff + A; where A is if (diff == vect [0]
[K]) A = vect [1] [k]; else A = vect [0] [k]; vect [2] [k + 1] = − diff + B: where B is if ((diff == vect [0 ]
[K]): :( diff == vector [1] [k]))
B = vector [2] [k]; else A = vector [1] [k]; vector [n] [k + 1] = − diff + P; where P is if ((diff == vect [0]
[K]): :( diff == vect [1]
[K]) :::: :( diff == vect [n-1]
[K]) P = vector [n] [k]; else P = vector [n−1] [k]; FIG. 5 shows an example of an apparatus for realizing the operation of the present embodiment.

In each of the embodiments described above, the compression method is realized by hardware, but this may be realized by software of a computer.

[Embodiment 3] In the above-described embodiment, when data does not match any vector, color data is attached after a code which does not match any vector. However, the present invention is not limited to this, and any information representing a color may be used. Absent.

For example, difference data may be attached, or a pallet code may be attached. Hereinafter, the pallet code will be briefly described.

When the fixed-bit palette information is attached, for example, if the 8-bit fixed palette is used, the number of colors is limited to 256. However, there is an advantage that only 8-bit data is required as compared with the case where 24-bit color data is attached. The palette may be determined first if the existing colors are known, or a new color that has appeared during compression may be registered. In this case, if it does not match any of the vectors, the registered colors in the palette are compared with the color data to be attached.If they match, the palette number is given.If they do not match, the new palette number is given, and the palette number is given. Output.

[Embodiment 4] In the above-described embodiment, the compression code output from the compression section may be further compressed by another lossless compression method. The other compression method is M of run-length compression method.
H or Huffman coding. Numerical value 1 from normal MH
Is preceded by 0 and the number is counted as one unit (for example, 0000
It is effective to perform Huffman coding as (100000000001). FIG. 8 shows the device of this embodiment.

[Embodiment 5] In the above embodiment, the compressed data are serially arranged and packed into one, but they may be packed separately.

For example, taking the first embodiment as an example, the information corresponding to code 0, the information corresponding to code 1, the information corresponding to code 2, and the information indicating the color to be attached may be separately stored in the memory (FIG. 9-901).

In the case of packing into one, since the place where the code data is stored is always undefined (for example, 32 bits
In the case where packing is performed at one depth, the code data may be stored from the first bit because the codes of code0, code1, and code2 are of variable length, and 5b.
(It may be an it-th time), a barrel shifter is required, and the packing circuit becomes large. This is solved by packing separately. Especially 24b
It is difficult to pack it color data, so code 0,
code1 and code2 may be stored together to separate only 24-bit color data (FIG. 9-902).

Also, code0, code1, code
2 are also separately packed, so that when the code data is further subjected to run-length compression, it often follows 0, and the effect of run-length compression is increased.

[Embodiment 6] In the above-described fourth embodiment, when the memory is divided into code 0, code 1, code 2, and color data, when the code data is subjected to run-length compression, the same data continues whether 0 or 1. Since the compression effect is higher, code0, code1, code2
May be toggled using the code assignment unit 211.

FIG. 10A shows reference numerals in accordance with the first embodiment. If this is converted into a toggle code, it becomes as shown in FIG. That is, the occurrence of 1 in FIG. 10A triggers the next sign to be inverted. As a result, 0s and 1s often follow, which is particularly effective when the code 0 is run-length later.

In the above embodiment, the flip-flop 2
Although the values stored in the vectors 06 and 210 are the values of the vectors that are highly likely to appear, a configuration as shown in FIG. 11 can be used. The counterpart to FIG. 2 is 20 in FIG.
6 is indicated by 1206 with the number 1000 added.
Here, the flip-flops 1206 and 1210 store not the vectors that are likely to appear, but information that can be used to calculate the vectors that are likely to appear by the operation units 1204 and 1208 using the information.

With respect to X and Y sequentially stored in FIG. 2 (FIG. 25 (c)), X 'and Y' stored in the apparatus example of FIG.
Is shown in FIG. FIG. 12- (a) shows the variation of X and Y stored in the flip-flop of the device of FIG. 2, and FIG. 12- (b) shows the variation of X 'and Y' stored in the device of FIG. It is a state of the mutation.

In FIG. 11, a vector having a high probability of occurrence (compared to X in FIG. 2) is calculated by the following computing unit 1204 so that X = −
X 'is calculated. Further, the vector having the next highest appearance (corresponding to Y in FIG. 2) is calculated as Y = − (X ′ + Y ′) by the calculators 1208 and 1212 to be described later. Therefore, the data stored in the flip-flop does not have to be the vector that has a high possibility of occurrence, and may be information that can be obtained by calculating a vector that has a high possibility of occurrence.

[0118]

As described above, according to the present invention, a multi-valued image created by a computer or the like can be compressed at a high compression ratio without deterioration. In particular, when differentially encoding a multi-valued image on a pixel-by-pixel basis, the value of the encoding target pixel is “it is likely to be the same value as the previous pixel”, and the value of the encoding target pixel is “two before. Paying attention to "there is a tendency to return from the pixel value appearing immediately before to the appearing pixel value", and applying 0 and x [k] corresponding to the above as targets of comparison with the difference value diff. Thus, efficient differential encoding can be performed.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an example of an overall system of a compression apparatus according to an embodiment.

FIG. 2 is a detailed view showing an example of a compression device.

FIG. 3 is a diagram illustrating an example of an entire system of a decompression device according to the embodiment.

FIG. 4 is a detailed view showing an example of a decompression device.

FIG. 5 is a detailed view showing an example of a compression apparatus according to another embodiment (1).

FIG. 6 is a flowchart showing the operation of the compression device of FIG. 2;

FIG. 7 is a flowchart showing the operation of the decompression device of FIG.

FIG. 8 is a diagram illustrating an example of a compression device system according to another embodiment (3).

FIG. 9 is a diagram illustrating an example of a compression device system according to another embodiment (4).

FIG. 10 is a diagram illustrating an example of a compression device system according to another embodiment (5).

FIG. 11 is a detailed view showing an example of a compression apparatus according to another embodiment (6).

FIG. 12 is a diagram showing changes in information X ′ and Y ′ in another embodiment (7).

FIG. 13 is an explanatory diagram of an ADCT compression method which is a conventional technique.

FIG. 14 is an explanatory diagram of an ADCT compression method which is a conventional technique.

FIG. 15 is an explanatory diagram of an ADCT compression method which is a conventional technique.

FIG. 16 is an explanatory diagram of an ADCT compression method which is a conventional technique.

FIG. 17 is an explanatory diagram of an ADCT compression method which is a conventional technique.

FIG. 18 is an explanatory diagram of an ADCT compression method which is a conventional technique.

FIG. 19 is an explanatory diagram of an ADCT compression method which is a conventional technique.

FIG. 20 is an explanatory diagram of an ADCT compression method which is a conventional technique.

FIG. 21 is an explanatory diagram of an ADCT compression method which is a conventional technique.

FIG. 22 is an explanatory diagram of an ADCT compression method which is a conventional technique.

FIG. 23 is an explanatory diagram of an ADCT compression method which is a conventional method.

FIG. 24 is a diagram showing a typical computer-generated image, a state of change of vectors X and Y, and compressed data.

FIG. 25 is a diagram illustrating an example of rearrangement performed by a pixel data rearrangement unit.

[Explanation of symbols]

 102 Compression unit 103 Data pack unit

Claims (4)

(57) [Claims] An input step of inputting pixel data d [i] in units of one pixel, the pixel data d [i], and pixel data d before the data
A difference output step of outputting a difference diff from [i-1]; a first determination step of determining whether the difference diff is 0; and determining whether the difference diff is x [k]. A second determining step of determining; and a code output step of outputting a code indicating the d [i] based on a match / mismatch in the determination by the first and second determining steps. An image data compression method, wherein when the difference diff is other than 0, x [k + 1] =-diff is updated and applied to the encoding of the next pixel data d [i + 1]. 2. An input step of inputting pixel data d [i] in units of one pixel, said pixel data d [i], and pixel data d before said data.
A difference output step of outputting a difference diff from [i-1]; a first determination step of determining whether the difference diff is 0; and determining whether the difference diff is x [k]. A second determination step of determining; a third determination step of determining whether the difference diff is y [k]; and a determination of the d based on a match / mismatch in the determination in the first to third determination steps A code output step of outputting a code indicating [i], wherein when the difference diff is not 0 in the first determination step, x [k + 1] = − diff, y [k + 1] = dif
f + A [k] (A [k] is the difference diff is x
Y [k] when matched with [k], x otherwise
[K]), and the next pixel data d [i +
An image data compression method, which is applied to the encoding of [1]. 3. An input means for inputting pixel data d [i] in units of one pixel, said pixel data d [i], and pixel data d preceding said data.
A difference output unit that outputs a difference diff from [i−1], a first determination unit that determines whether the difference diff is 0, and whether the difference diff is x [k]. A second determining means for determining; and a code output means for outputting a code indicating the d [i] based on a match / mismatch in the determination by the first to third determining means. When the difference diff is other than 0,
An image data compression apparatus for updating x [k + 1] =-diff and applying the update to the encoding of the next pixel data d [i + 1]. 4. An input means for inputting pixel data d [i] in units of one pixel, said pixel data d [i], and pixel data d preceding said data.
A difference output unit that outputs a difference diff from [i−1], a first determination unit that determines whether the difference diff is 0, and whether the difference diff is x [k]. A second determining means for determining, a third determining means for determining whether or not the difference diff is y [k], and a d based on a match / mismatch determined by the first to third determining means. Code output means for outputting a code indicating [i], and when the difference diff is other than 0 in the first determination means,
x [k + 1] = − diff, y [k + 1] = diff +
A [k] (A [k] is set to y [k] when the difference diff matches x [k] and x [k] otherwise), and the next pixel data d is updated. An image data compression apparatus, which is applied to encoding of [i + 1].