JPH0654208A - Image processing method/device - Google Patents

Abstract

PURPOSE:To compress the image data with high picture quality and at the fixed compression rate by selecting the reversible compression processing or the non-reversible compression processing based on the relationship between the quantity of compressed data undergone the reversible compression of a block and a prescribed quantity of data. CONSTITUTION:An image processor which processes the image data for each block is provided with a reversible compressing means 3 which reversibly compress the blocks and an ADCT compressing means 4 which non-reversibly compress the image data. Then the block undergoes the reversible compressing processing with D<=L and the non-reversible compression processing with D>L respectively (D: the compressed data quantity obtained with the reversible compression of the block; L: the prescribed data quantity). Then the result obtained by expanding the compressed data as the reversible compressed data or the result obtained by expanding the compressed data as the non-reversible compressed data is selected baded on the input information. Thus, the original image data are restored. As the result, the images produced by a computer can be reversibly compressed and the picture quality of the natural images can be improved with the non-reversible compression processing.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method and an image processing apparatus, for example, an image compression method, an image compression apparatus, and an image compression method for switching image data by switching between lossless compression and lossy compression according to the type of image. , A data decompression method for decompressing data in which lossless compression data and lossy compression data are mixed.

[0002]

2. Description of the Related Art In recent years, with the spread of DTP and the like, it has been required to improve the image quality of an image created on a computer, and colorization and multi-gradation have been advanced. The amount of information in this type of image is, for example, about 46 Mbytes in the case of A4 size, 400 dpi, 256 gradations, and three primary color colors.
If the image data is handled as code information such as a page description language (hereinafter referred to as “PDL”), the amount of information is reduced, but it takes time to develop the code information into image data, and the code information is developed. However, there is a problem that the original image data cannot be completely reproduced, and therefore the image data compression technique becomes important.

As a general compression method for color multi-valued images, ADCT (Adaptive Discrete Cosine Transform) recommended by JPEG (Joint Photographic Expent Group)
There is a method, and the ADCT image compression method will be described below. FIG. 14 is a block diagram showing the configuration of the ADCT image compression device. In FIG. 14, a color conversion unit 3101 converts the RGB color space of the NTSC system into a YCrCb color space represented by a luminance signal Y and two color difference signals Cr and Cb.

A sub-sampling unit 3102 reduces the color difference data by utilizing the characteristics that human eyes are sensitive to luminance and insensitive to color difference. Specifically, the sub-sampling unit 3102 reduces the amount of color difference data by half by taking the average value of two adjacent color difference data. Three
Reference numeral 103 denotes a DCT, which divides the image data output from the sub-sampling unit 3102 into horizontally and vertically adjacent 8 × 8 blocks, and further performs DCT (discrete cosine transform) to transform the frequency space data. To do.

A quantizer 3104 divides each 8 × 8 = 64 DCT coefficients by a quantized value having a different step width to obtain quantized data. A coding unit 3105 divides the 64 quantized DCT coefficients output from the quantization unit 3104 into one DC coefficient and 63 AC coefficients, and codes each coefficient according to the Huffman table recommended by JPEG. Turn into. The encoded data is attached to a header such as quantization table data and Huffman table data, and then stored in a memory or transmitted to another device.

FIG. 15 is a block diagram showing the structure of the ADCT data decompression device. In FIG. 15, a decoding unit 3205 decodes the input code data into quantized data. An inverse quantization unit 3204 converts the quantized data output from the decoding unit 2305 into DCT coefficient data. This can be obtained by multiplying each of the 64 coefficients by the quantized value using the quantization table used when the quantizer 3104 of FIG. 14 quantizes.

An inverse DCT 3203 is an inverse quantizer 320.
Inverse DCT is applied to the DCT coefficient data output from No. 4 to obtain actual image data. 3202 is an interpolation unit, which is an inverse DC
C missing in the image data output from T3203
The r and Cb data are simply interpolated by an iterative method. The missing of Cr and Cb data occurs when the sub-sampling unit 3102 of FIG. 14 compresses the data.

A color conversion unit 3201 converts the YCrCb color space data output from the interpolation unit 3202 into NTSC-R.
Convert to GB color space data or color space data of the device. Next, the processing flow of the ADCT image compression apparatus will be described by showing actual data. FIG. 16 shows a part of image data of a color multi-valued image created by a computer, which is NTSC-RGB color space data, and a part (1) of the image data in which the character part of the color multi-valued image is developed.
6 pixels × 16 pixels).

In FIG. 16, 3301 is R data, 3
302 indicates G data and 3303 indicates B data. In addition,
The data range of each pixel is 8 bits (0 to 255),
In the figure, a part of the blue character of (R, G, B) = (30,30,225) is drawn on a slightly dark white background of (R, G, B) = (225,225,225). The color conversion unit 3101 uses the following equation:
Perform a conversion from the NTSC-RGB color space to the YCrCb color space.

Y = 0.299 × R + 0.587 × G + 0.114 × B Cr = 0.713 (RY) Cb = 0.564 (BY) Further, according to the CCIR recommendation, overshoot and undershoot at the time of calculation in the YCrCb color space are performed. To allow, the data is rounded according to the following formula.

Y = 219.0 × Y + 16.5 Cr = 224.0 × Cr + 128.5 Cb = 224.0 × Cb + 128.5 The sub-sampling unit 3102 performs sub-sampling on the Cr and Cb data obtained by the above equations. As the sub-sampling method, simple thinning, MAX
Although there are data selection, MIN data selection, etc., an example using the average value method will be described here. That is, the average value method is a method in which the average value of the data of two adjacent pixels is taken as one data.

FIG. 17 is a diagram showing a result of processing the image data shown in FIG. 16 by the color conversion unit 3101 and the sub-sampling unit 3102. In FIG. 17, 3401 is Y data, 3402 is Cr data, 3403 is Cb data, and the amount of each data of Cr and Cb is reduced to 1/2 by subsampling.

Next, the Y, Cr, Cb data shown in FIG. 17 is input to the DCT 3103, and the DCT 3103
First, the input data is horizontally and vertically adjacent to each other 8
Divide into blocks of × 8 data. Due to blocking,
As shown in FIG. 17, the Y data 3401 is 3401a.
~ 3401d divided into 4 blocks. Similarly, Cr data 3402 is divided into 2 blocks 3402a and 3402b, and Cb data 3403 is divided into 2 blocks 3403a and 3403b.
It is divided into blocks. Subsequently, the DCT 3103 executes DCT on each of these eight blocks.

FIG. 18 is a block diagram of the eight blocks shown in FIG.
It is a figure which shows a mode that T-transformed. In FIG. 18, 35
01 indicates the DCT coefficient of the Y data, and the block 3501
a to 3501d are blocks 3 shown in FIG.
It corresponds to 401a-3401d. Similarly, 3502 is the DCT coefficient of Cr data, and 3503 is the Db of Cb data.
The CT coefficient is shown. Each block after DCT is composed of one DC component in the upper left corner and 63 other AC components.

Next, the DCT coefficient data 35 shown in FIG.
01 to 3503 are quantized by the quantization unit 3104.
The quantization here is performed using a quantization table recommended by JPEG. FIG. 19 is a diagram showing a quantization table used for quantization. In FIG. 19, 4001 is for the Y component, 4
002 is a quantization table for Cr and Cb components.

FIG. 20 is a diagram showing the result of quantizing the DCT coefficient data shown in FIG. 18 with the quantization table shown in FIG. In FIG. 20, 3601 is Y component quantized data, 3602 is Cr component quantized data, 306
3 is the quantization day of the Cb component. The code part 3105 is
The quantized data 3601 to 3603 shown in FIG.
The C component and the AC component are divided, and for the DC component, an optimum Huffman coding table is created from the histogram of the difference from the DC component of the previous block, and the optimum Huffman coding table is coded according to the table. , After performing sorting in the zigzag order as shown in FIG. 21, the optimum Huffman coding is performed from the histogram of the combination of the run length of the coefficient 0 (the number of consecutive coefficients 0 until the coefficient x other than 0 appears) and the value x. Create a table and encode according to the table.

Since 16 × 16 pixel data in NTSC-RGB has a range of 8 bits per pixel, the original image data of 16 × 16 × 3 colors × 8 bits = 6,144 bits is the code part. When output from 3105, it is compressed to 795 bits. Therefore, the image data is compressed to 6,144 / 795≅7.7 approximately 1 / 7.7. However, in reality, since the image size, the quantization table, the encoding table, and the like are attached to the data encoded by the encoding unit 3105, the compression rate is slightly reduced.

Next, the processing flow of the ADCT data decompression device will be described. The decoding unit 3205 receives the input ADCT
Decode the compressed image data. The inverse quantization unit 3204 performs inverse quantization by multiplying the quantized data output from the decoding unit 3205 by the coefficient of the quantization table shown in FIG. The data shown in FIG. 22 is obtained by the above processing. FIG. 22 is a diagram showing the result of Huffman decoding and inverse quantization performed on ADCT compressed image data. 3701 represents the Y component, 3702 represents the Cr component, and 37.
03 shows a Cb component.

If the dequantized data (DCT coefficient data) shown in FIG. 22 and the DCT coefficient data before quantization shown in FIG. 18 are compared, it is clear that the respective data are different. Then, the inverse DCT 3203 converts the DCT coefficient data output from the inverse quantizer 3204 into an inverse D
CT and return to YCrCb data. 23 is a diagram showing the result of inverse DCT of the DCT coefficient data shown in FIG. 22 by the inverse DCT 3203. 3801 indicates Y data, 3802 indicates Cr data, and 3803 indicates Cb data.

Subsequently, the interpolation section 3202 determines the inverse DCT 32.
03 interpolates the missing data of the image data output,
The color conversion unit 3201 outputs the Y output from the interpolation unit 3202.
The CrCb color space data is converted into NTSC-RGB color space data. FIG. 24 is a view showing the data finally obtained by the ADCT data decompression device, 3901 indicates R data, 3902 indicates G data, and 3903 indicates B data.

[0021]

However, the above conventional example has the following problems. That is, the ADCT image compression method is a lossy compression method that involves data loss during subsampling and quantization. Therefore, FIG.
24, which is the result of compressing and expanding the NTSC-RGB color space data before compression shown in FIG. 16 and the data of FIG. 16 shown in FIG.
As is clear from comparison with the TSC-RGB spatial data, in the ADCT image compression method, the decompressed data is different from the uncompressed data, that is, there is a drawback that image quality deterioration occurs. It was

Further, the image created on the computer (computer created image) has advantages such as a clean outline and noise-free color painting by one figure (or one character) single color. However, when the image data is compressed and expanded by the ADCT image compression method, the outline of the figure is disturbed, pseudo-edges called mosquito noise are generated, and coloring is caused by quantization, so that the advantage of the computer-generated image is not exhibited. In particular, since the above-mentioned compression method is an 8 × 8 block process, there is a drawback that a large color change occurs in the boundary area.

If an attempt is made to further increase the compression rate by the ADCT image compression method, finally, the AC of the DCT coefficient will be
The component is lost and block distortion occurs. For example, the resolution is 1
However, there is a drawback that extreme deterioration of resolution occurs such as / 8. On the other hand, in order to improve the image quality, there is a method of compressing the image data by a lossless compression method such as Huffman coding, but there is a drawback that the amount of data after compression changes. Reversible compression of an image or a natural image read by an image scanner may result in an extremely large amount of data, and it is necessary to prepare an image memory having a large storage capacity assuming the worst case.

[0024]

SUMMARY OF THE INVENTION The present invention is intended to solve the above problems, and has the following structure as one means for solving the above problems. That is, in the image processing method for processing the image data in block units, the compressed data amount D when the block is losslessly compressed.
When the relationship between the predetermined value L and D is L ≦ L, the block is losslessly compressed, and when the relationship between the compressed data amount D and the predetermined value L when the block is losslessly compressed is D> L. An image processing method for irreversibly compressing the block is used.

An image processing apparatus for processing image data in block units, which comprises a first compression means for reversibly compressing the block and a second compression means for irreversibly compressing the block.
Compression means, a comparison means for comparing the compressed data amount D when the block is compressed by the first compression means with a predetermined value L, and a comparison result of the comparison means is D ≦ L. Selecting the compression result of the first compression unit, and selecting the compression result of the second compression unit when the comparison result of the comparison unit is D> L, the comparison result of the comparison unit And the compression result selected by the selecting means.

Also, in the image image processing method for decompressing the compressed data obtained by compressing the image data block by block, the result of decompressing the compressed data as reversible compression data and the decompression process of the compressed data as irreversible compression data. One of the results is selected based on the input information, and the selected decompression results are combined to provide the image processing method for restoring the original image data.

[0027]

With the above configuration, when the relationship between the compressed data amount D and the predetermined value L when the block is losslessly compressed is D≤L, the block is losslessly compressed and D> L. In this case, the block is lossy-compressed, and one of the result of decompressing the compressed data as lossless compressed data and the result of decompressing the compressed data as lossy-compressed data is based on the input information. An image processing method and an image processing apparatus for selecting and restoring original image data can be provided.

For example, with the above configuration, the image data created by the computer is reversibly compressed as much as possible to prevent the deterioration of the image quality, and the lossless compression of the gradation image and the natural image is sufficient. It is possible to prevent the memory shortage by performing efficient compression by performing lossy compression on the image data for which the rate cannot be expected.

The image data for which the compression ratio cannot be expected by the lossless compression method is often a natural image in which the image data varies. Therefore, according to the expression D> L, there is a considerable probability that the computer-created image and the natural image are separated into regions. It is possible. Since natural images have the characteristic that even if the data changes due to ADCT compression, the deterioration is not felt by human eyes, this method enables compression with good image quality and a constant compression rate. Become.

[0030]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image processing apparatus and a data decompression apparatus according to an embodiment of the present invention will be described in detail below with reference to the drawings.
It should be noted that in the following explanation, the ADCT
An example will be described in which the block size is 8 × 8 pixels and the pixel data is 24 bits, but the present embodiment is not limited to this.

In the following description, an example in which the original image data is NTSC-RGB 24 bits will be described, but the present embodiment is not limited to this. For example, CMYK, XYZ, CIE-Lab, CIE-Lu
v, YCrCb may be used, data may be 16 bits and 36
Bits are fine. In the following description, "area 1" is an area for reversibly compressing blocks, and "area 2" is
Is an area for irreversibly compressing the block.

[0032]

[First Embodiment] FIG. 1 is a block diagram showing a configuration example of an image compression apparatus according to the present embodiment. In FIG. 1, reference numeral 1 denotes an image data storage unit which outputs, for example, 24-bit / pixel data in the NTSC-RGB format in block units. A delay unit 2 holds the block output from the image data storage unit 1 and delays it by, for example, one block in order to determine execution / non-execution of ADCT compression based on the result of lossless compression.

A lossless compression unit 3 performs lossless compression of the block output from the image data storage unit 1 and region determination of the block, and outputs compressed data and a signal S2g representing the region determination result. An ADCT compression unit 4 performs ADCT compression on the data input from the delay unit 2 when the region determination result of the lossless compression unit 3 is “region 2”, and the compressed data is a DC component, a low frequency component, and a slightly high frequency component. Divide into components, high frequency components, etc. and output.

A buffer 5 stores the compressed data output by the lossless compression unit 3, and outputs the stored compressed data when the region determination result of the lossless compression unit 3 is "region 1". A segment controller 6 stores the compressed data from the buffer 5 and the compressed data from the ADCT compression unit 4 in the first storage unit 8. Further, if the first storage unit 8 becomes full during the compression process, the segment controller 6 refers to the segment information table 7 and stores ADCT compressed data of lower importance, for example, higher frequency side ADCT compressed data. Overwriting of compressed data is prevented by overwriting the address with new compressed data.

A segment information table unit 7 stores the storage address of the data stored in the first storage unit 8 by the segment controller 6. 9 is a second storage unit,
A signal S representing the area determination result output from the lossless compression unit 3.
Remember 2g. A control unit 10 controls the operation timing of this embodiment including the above-described configurations.

With the above configuration, the first storage unit 8 shown in FIG. 1 stores divided ADCT compressed data in accordance with the signal S2g representing the region determination result output from the lossless compression unit 3, or Lossless compressed data is stored. FIG. 2 is a block diagram showing a configuration example of the reversible compression unit 3. In FIG. 2, a rearrangement unit 201 rearranges the input pixel data of the blocks in serial.

A first latch unit 202 latches pixel data (for example, 24 bits) from the rearrangement unit 201. Similarly, 203 to 205 are second to fourth latch units, which latch the pixel data output from the first latch unit 202, respectively. A first comparing unit 206 compares the pixel data output from the rearranging unit 201 with the pixel data latched by the first latching unit 202, and when both are equal, it is "0".
If the two are different, a 1-bit signal S that becomes "1"
Output 2a. That is, the signal S2a represents whether or not the color represented by the pixel data serially output from the rearrangement unit 201 has changed, and becomes “1” when the color has changed.

Similarly, reference numerals 207 to 209 denote second to fourth comparison units, which compare the pixel data output from the rearrangement unit 201 with the pixel data latched by the corresponding latch units 203 to 205. , And sends the comparison result to the decoder 210. Two
A decoder 10 receives the comparison results of the second to fourth comparing sections 207 to 209 and outputs a 2-bit signal S2b.
The signal S2b is "00" when the second comparison unit 207 matches and "01" when the third comparison unit 208 matches.
Further, when the fourth comparing unit 209 agrees, it becomes “10”, and when all do not coincide, it becomes “11”. That is,
The signal S2b indicates where in the latch units 203 to 205 the data having the same value (same color) as the pixel data output from the rearrangement unit 201 has been latched, and when it is not latched anywhere, it is '11. 'become.

Reference numeral 211 denotes a PLD (Programmable Logic Devic
In e), the signal S2a input from the first comparison unit 206, the signal S2b input from the decoder 210, and the pixel data signal S2c input from the rearrangement unit 201 are input, the compressed data of the lossless compression unit 3, and 2nd-4th latch parts 203-20
The latch signals S2d to S2f are output to 5. Where PLD
Is an IC that can be programmed by the designer.

The compressed data output by the PLD 211 has the relationship shown in Table 1 as an example.

[0041]

[Table 1] * 1: 24-bit color data Table 2 shows the input of PLD211 and the latch signals S2d to S2.
In the table showing the relationship of 2f, when the latch signal is "1", each corresponding latch unit latches the input data.

[0042]

[Table 2] As shown in Table 2, when the signal S2a is "0" (the color has not changed), any latch signal becomes "0" regardless of the state of the signal S2b, and the signal S2a becomes "1". In the case of "(color has changed)" and the signal S2b is "11" (pixel data of the same color has not been latched), the latch part in which the oldest pixel data is latched is rewritten.

Reference numeral 214 denotes an area determination unit, which is the input signal S
According to 2a and signal S2b, a 1-bit signal S2g representing the determination result is output. The state of the signal S2g is "0",
For "1", the judgment results are "area 1" and "area 2", respectively.
It means that A backup unit 213 stores the pixel data latched in each latch unit before the determination of the region determination unit 214, and after the determination, when the signal S2g is '1', the regions to each latch unit are stored. Returns the pixel data before judgment.

FIG. 3 is a block diagram showing an example of the structure of the area determination unit 214. In FIG. 3, 312 to 316 are registers. 311 is an AND gate, and the signal S
The logical product of 3 bits in total of 2a and the signal S2b is output. Thirty
Reference numeral 1 denotes a counter a, which counts the number of times the signal S2a is "1", that is, the number of times a color change point appears.

A counter b is an AND gate 31.
The number of times the output of 1 is "1" is counted. That is, the counter b302 counts the number of times the signal S2a is "1" and the signal S2b is "11", that is, the number of times a color that has not been latched anywhere appears. Reference numerals 303 and 304 denote multiplication units. The multiplication unit a303 multiplies the output of the counter a301 by 2 (the number of bits of the signal S2b) stored in the register 312, and the multiplication unit b304 outputs the output of the counter b302 and the register 313. 24 (the number of bits of pixel data) stored in the memory.

Reference numeral 305 denotes an adder a which adds the output of the multiplier a303, the output of the multiplier b304, and 64 (the number of color change point information) stored in the register 314, and finally adds the block. Number of bits required by (compressed data amount)
Output D. The number of color change point information is a value determined by the set block size, and is not limited to 64.

A subtraction unit 306 outputs a result CD obtained by subtracting the output of the addition unit a305 from the value C stored in the register 315. The value C is the target data size per block after compression. For example, if the target compression rate in this embodiment is 1/12, the value will be as follows from the block size and the pixel data size. C = 8 × 8 × 24/12 = 128 bytes 307 is an addition unit b, which is a subtraction result CD of the subtraction unit 308.
And the output S of the latch 309 described later are added, the result C +
Output S-D.

A comparison unit 308 compares the addition result C + S−D of the addition unit b307 with the value 0 stored in the register 316, and outputs a signal S2g representing the region determination result.
The signal S2g is "0" if C + S-D≥0,
If C + S-D <0, it becomes "1", and as described above, the states "0" and "1" of the signal S2g indicate that the determination results are "region 1" and "region 2", respectively. .

That is, if the target compression ratio can be achieved by adding a savings S (described later) after block compression (C + S-D ≧ 0), the region determination unit 214 determines that the block is a “region 1” for lossless compression. If the savings S described later is insufficient to achieve the target compression ratio (C + S−D <0), it is determined that the block is the “region 2” in which the block is irreversibly compressed. 310
Is a latch, which is initialized to a predetermined value by the control unit 10 at the start of the compression process, and thereafter, when each block is determined to be “area 1”, that is, when the signal S2g is “0”, addition is performed. The output C + S-D of the part b307 is latched. That is, the latch 310 latches the margin C + S-D (> 0) with respect to the target compression rate at that time after one block compression, and outputs the margin S.

In this embodiment, 128 is set as the initial value of the margin S, but 0 is not set in the margin S at this time when a large number of different colors appear in the first block. This is because. This is because the compression method of the lossless compression unit 3 requires 64 bits per block for the information indicating the color change point and 2 bits for the latch information at the color change point, and the new color that is not latched is used. When 24 appears, 24 bits are required to output the data of that color as it is, and the number of changes in color is n, and the number of appearance of a new unlatched color is m.

D = 64 + 2n + 24m ∴C + S− (64 + 2n + 24m) ≧ 0 However, 0 ≦ n ≦ 64, 0 ≦ m ≦ n Therefore, in the compression method of the reversible compression unit 3, C = 128, S = in the first block. 0, m = n, n <(128-64) /26=2.5 That is, since the upper limit is 2 colors, it is necessary to prepare some margin S first, for example, C = S
= 128, n <7.4, and when m = n, compression is possible even when up to 7 colors appear.

FIG. 4 is a block diagram showing a configuration example of the ADCT compression unit 4. In FIG. 4, a color space conversion unit 801 converts RGB color space data into YCrCb color space data. A DCT unit 802 performs DCT on the YCrCb data output from the color space conversion unit 801, and converts the YCrCb data into frequency space data.

A quantizing unit 803 quantizes the DCT coefficients output from the DCT unit 802 with different step widths. Reference numeral 804 denotes a coding unit, which is a quantization unit 803.
The quantized data output from is divided into a DC coefficient and an AC coefficient, and with respect to the DC coefficient, an optimal Huffman coding table is created from the histogram of the difference from the DC coefficient of the previous block, and according to the table, Encoding and also A
Regarding C coefficient, after rearranging in zigzag order,
From the histogram of the combination of the run length of the coefficient 0 (the number of consecutive coefficients 0 until the coefficient x other than 0 appears) and the value x,
An optimum Huffman coding table is created, and coding is performed according to the table. The encoding unit 804 outputs the code data of the DC coefficient to the line 8a, and the AC coefficient of the p coefficient code data to the line 8b, the q code data to the line 8c, and the r coefficient data to the r coefficient, respectively. The code data is distributed to the lines 8d, ...

The signals S2a, S2b and S2c can be stored in different addresses in the first storage section 8 by the segment controller 6. Further, the latch portion that latches the past pixel data may be one, that is, the first latch portion and the second latch portion. 12
FIG. 4 is a block diagram showing another configuration example of the lossless compression unit 3, which is a configuration example of the lossless compression unit 3 when there is one latch that stores past pixel data and the margin S is not set. in this case,
There are also two comparators, the signal S2b is one bit, and the formula for calculating the amount of compressed data is as follows.

D = 64 + count2 + 24 × count3 Further, in the lossless compression section 3, the latch part to be updated when no pixel data having the same value exists in any latch part may be fixed. For example, when the signal S2a is "1" and the signal S2b is "11", the second latch unit 203 is updated. In this case, the compression efficiency is slightly worse, but the hardware can be simplified.

Furthermore, in the above description, the segment controller 6 is used for memory control to perform data compression in real time. However, pre-scan (pre-expansion) is performed to preliminarily perform the delta factor of ADCT compression. The compression may be performed after the determination. In this case, memory control can be simplified. As described above, according to the present embodiment, the block that can achieve the target compression rate is compressed by the reversible compression unit 3, and the block that cannot achieve the target compression rate is AD.
Since it is compressed by the CT compression unit 4, the image quality of computer-created image data and the like is reversibly compressed as much as possible to prevent deterioration of the image quality, and a sufficient compression rate is expected in lossless compression such as gradation images and natural images. For image data that cannot be processed, ADCT compression can be used to perform efficient compression and prevent memory shortage.
Unlike the computer-generated image, the gradation image and the natural image have the characteristic that the image quality is not conspicuous even when the image is lossy compressed and expanded.

Furthermore, according to the present embodiment, lossless compression and lossy compression can be switched in block units. For example, even a natural image embedded in a computer created image can be embedded in a natural image. It also works effectively for computer created images. Further, according to the present embodiment, although the compression is performed according to the above procedure,
If there is still a shortage of memory for storing the compressed data, the lack of memory can be resolved by sequentially deleting the ADCT code data on the higher frequency side.

[0058]

[Second Embodiment] An image compression apparatus according to a second embodiment of the present invention will be described below. The second embodiment implements the lossless compression unit 3 of the first embodiment by software.
Therefore, only the lossless compression unit will be explained below,
Detailed description of other configurations that are substantially the same as those of the first embodiment will be omitted.

The contents of the data name in the following description are as follows, and the parentheses show the correspondence with the first embodiment. next: First latch data (first latch unit 202) former [0]: Second latch data (second latch unit 203) former [1]: Third latch data (third latch unit 204) former [2]: Fourth latch data (Fourth latch unit 205) present: Pixel data output in zigzag order (output of rearrangement unit 201) turn: Latch update information Each data of next, former [], present is RGB format pixel data. , R, G, B data respectively.

FIG. 5 is a flow chart showing an example of the lossless compression procedure of this embodiment. In FIG. 5, the present embodiment is
In step S401, the margin S is initialized to 128,
In step S402, the four latch data former [] are initialized. In the present embodiment, for example, assuming that the black image, the red image, and the blue image appear on a white background in descending order of frequency of occurrence, white data is written to next [former [0]]
Set the black data to, the red data to the former [1], and the blue data to the former [2].

Next = (255,255,255) former [0] = (0,0,0) former [1] = (255,0,0) former [2] = (0,0,255) If present does not match any former [] in S402, a turn for rewriting the oldest former [] is set as follows.

Turn = 0 turn_p = 1 turn_pp = 2 Next, in the present embodiment, in step S403, next, former
Back up each pixel data of [] and turn.

Bnext ← next Bformer [] ← former [] Bturn ← turn Bturn_p ← turn_p Bturn_pp ← turn_pp In the present embodiment, step S404 will be described later in detail, but one block is reversibly compressed to obtain compressed data. Find the quantity D.

Subsequently, in the present embodiment, step S405 is executed.
Then, the compressed data amount D obtained in step S404 is compared with a limit value L obtained by adding a margin S to a preset value C, and if D≤L, it is determined to be "area 1" and step S4 is performed.
If it is D> L, it is determined to be "region 2" and the process proceeds to step S409. If the determination result is "area 1", in this embodiment, the state is "0" in step S406.
The signal S2g is output, the margin S is updated by the following equation in step S407, and then the process proceeds to step S410.

S ← C + SD If the judgment result is "area 2", in this embodiment, the signal S2g of the state "0" is output in step S408, and next, former [] are output in step S409. , After restoring the backed up pixel data to turn, step S
Proceed to 410. That is, in this embodiment, step S40
In step 9, each of the latch data rewritten in the process of step S404 is returned to the pixel data before the process.

Next ← B next former [] ← Bformer [] turn ← Bturn turn_p ← Bturn_p turn_pp ← Bturn_pp Then, in this embodiment, in step S410, it is determined whether or not all the blocks have been processed, and the processing is not completed. If there is any block, the process returns to step S403 to process the remaining blocks, and if there is no unfinished block, the process ends.

Next, the processing procedure for lossless compression (step S404) will be described. The contents of the data name in the following description are as follows, and the parentheses indicate the correspondence with the first embodiment. count2: the number of times a color change point appears (counter a301) count3: the number of times that it does not match any latch data (counter b302) pix [i]: i-th pixel data FIGS. 6 and 7 show an example of the processing procedure of step S404. It is a flow chart shown.

Referring to FIG. 6, this embodiment uses step S5.
Initialize count2 and count3 to 0 with 01, and step S50
At 2, the counter i is set to 0. The counter i counts the number of pixels to be processed, and the present embodiment processes pixels in the order shown in FIG. Subsequently, in this embodiment, pix [i] is substituted into present in step S503,
In step S504, the present and the next are compared. If they match, the compressed data '0' is output in step S505, and if they do not match, the details will be described later in step S505.
The region 2 processing of 506 is executed. That is, in this embodiment, in step S504, the current pixel data (presen
Since t) is compared with the immediately preceding pixel data (next),
A match between both data indicates that there is no color change, and a mismatch between both data indicates that there is a color change.

Subsequently, in the present embodiment, step S507 is executed.
The latched pixel data is assigned to next, the counter i is incremented in step S508, the value of the counter i is determined in step S509, and if i ≦ 63, the process returns to step S503 to process the remaining pixels. Then
If i> 63, the process proceeds to step S510. When the processing of all pixels is completed (i> 63), in this embodiment, in step S510, the compressed data amount D of the block is
Is obtained by the following equation, the process returns to the procedure shown in FIG.

D = 64 + 2 × count2 + 24 × count3 On the other hand, if present ≠ next in step S504, the processing procedure shown in FIG. 7 is executed. In FIG. 7, in the present embodiment, in step S601, the compressed data "1" is set.
Is output, count2 is incremented in step S602, turn and turn_p are compared in step S603, and if they match, jump to step S605. If they do not match, step S604 follows and turn_pp
To turn_p after substituting the turn_p latch data number
Substitute the latch data number of. This is turn and tu
This is a process for not giving the same value to rn_p.

Then, in this embodiment, step S605 is executed.
Then, compare present and former [], present = former [0]
If so, it branches to step S606 if present = former [1], to step S610 if present = former [2], and to step S612 if present does not match any former []. present = former
In the case of [0], in this embodiment, the compressed data '00' is output in step S606, and the turn data is set to '0' in step S607 to indicate the latch data to be updated.

If present = former [1], in this embodiment, the compressed data '01' is output in step S608, the turn data '1' is set in step S609, and the updated latch data is shown. If present = former [2], the present embodiment outputs the compressed data '10' in step S610, sets '2' to turn in step S611, and indicates the updated latch data.

If present does not match any former [], the present embodiment outputs the compressed data '11' in step S612, outputs the pixel data latched in present in step S613, and outputs the step. In S614
Increment count3, and in step S615 turn_p
And a turn data number different from turn_pp are set to turn. In addition, in steps S603 and S604 described above, tu
Since rn_p and turn_pp are not set to the same value, turn is either 0, 1, or 2.

Then, in this embodiment, in step S616.
After substituting the pixel data that is latched in next into former [turn], the procedure returns to the processing procedure shown in FIG. It should be noted that since the number of colors used is relatively small in a computer-generated image in DTP or the like, and 256 colors are generally sufficient, the present value present output in step S613 described above can also be converted into a lookup table. is there. Also, for example,
A memory of 8 bits (256 colors) is prepared as a table, the compression rate is improved by outputting data of 8 bits / pixel up to the 256th color, and after the 256th color appears,
By outputting 24-bit / pixel data as it is, the number of colors is not limited.

The current value output in step S613
It is possible that the present is not 24 bits but 18 bits, for example. In this case, it is not lossless compression, but in high-frequency images such as computer-generated images,
The deterioration is not noticeable because the gradation to the human eye is not important. The amount of compressed data in this case is as follows. D = 64 + 2 × count2 + 18 × count3 Further, one former may be used to latch past pixel data. In this case, steps S606, S608, S
The signal output at 610 and S612 is 1 bit,
The latch to which the latched pixel data is assigned to next is 1
Therefore, the turn indicating the update order is unnecessary. In this case, the formula for calculating the amount of compressed data is as follows.

D = 64 + count2 + 24 × count3 As described above, according to the present embodiment, in addition to the effect similar to that of the first embodiment, a reversible compression section is constructed by software to perform reversible compression. In addition, since the compressed data amount D is obtained, the hardware can be simplified.

[0077]

[Third Embodiment] Next, a data decompression apparatus according to a third embodiment of the present invention will be described. In the third embodiment, the compressed data stored in the first storage unit 8 and the second storage unit 9 shown in FIG. 1 of the first embodiment is expanded to reproduce the image data. FIG. 9 is a block diagram showing a configuration example of the data decompression device of this embodiment.

In FIG. 9, reference numeral 901 denotes a segment controller which refers to the segment information table 7 shown in FIG. 1 to convert the compressed data stored in the first storage unit 8 into an AD
It is distributed to the CT decompression unit 902 or the lossless compressed data decompression unit 904. An ADCT decompression unit 902 decompresses the irreversibly compressed data and reproduces a block of 8 × 8 pixels.

Reference numeral 904 denotes a lossless compression data decompression unit (hereinafter simply referred to as “decompression unit”) which decompresses the losslessly compressed data based on the area information stored in the second storage unit 8
Play a block of × 8 pixels. 906 is a selector,
Based on the area information stored in the second storage unit 9, the ADC
The block input from the T extension unit 902 or the extension unit 90
One of the blocks input from 4 is selected and output.

Reference numeral 907 denotes an image data storage unit, which is a selector 9
Blocks output from 06 are sequentially stored and, for example, N
Image data of 24 bits / pixel in TSC-RGB format is reproduced. A control unit 910 controls the operation timing of the present embodiment including the above-described configurations.

FIG. 10 is a flow chart showing an example of a processing procedure for realizing the lossless compression data decompression unit 904 by software. The data names are substantially the same as those in the second embodiment. In FIG. 10, this embodiment is based on step S100.
In step 1, next, former [], and turn are set to the step S of the second embodiment.
Initialization is performed in substantially the same manner as 402.

Then, in this embodiment, step S1002 is performed.
If the area information of the processing block read from the second storage unit 9 indicates "area 1", the decompression processing, which will be described in detail later, is executed in step S1003, and then step S100.
4 and if "region 2" is indicated, step S1
Jump to 004. Subsequently, in the present embodiment, in step S1004, it is determined whether or not all the compressed data have been finished, and if there is unprocessed compressed data, step S10.
Returning to step 02, the processing is ended when the expansion of all the compressed data is completed.

FIG. 11 is a flow chart showing the procedure of the extension processing of step S1003. In FIG. 11, in the present embodiment, the counter i is set to 0 in step S1101. It should be noted that the counter i counts the number of pixels to be processed, and this embodiment is similar to the second embodiment in FIG.
The pixels are processed in the order shown in FIG. Then, in the present embodiment, in step S1102, the 1-bit data received from the first storage unit 8 is determined, and if the data is "1", the process proceeds to step S1103, and the data is "0". If ', then jump to step S1113.

If the one-bit data received from the first storage unit 8 is "1", the present embodiment is step S11.
In 03, turn and turn_p are compared, and if they are different, jump to step S1105. If they are different, the process proceeds to step S1104 to turn_pp to turn_pp.
After assigning the latch data number of p, assign the latch data number of turn to turn_p. This is a process for not giving the same value to turn and turn_p.

Then, in this embodiment, step S1105 is executed.
Then, depending on the 2-bit data received from the first storage unit 8, in the case of "00", go to step S1106 and "01".
If it is, go to step S1107, if it is '10', go to step S1108, and if it is '11', go to step S1108.
Branch to 1110. If the 2-bit data received from the first storage unit 8 is "00", this embodiment is
At 1106, after 0 is substituted for turn, step S110 is executed.
Proceed to 9.

When the 2-bit data received from the first storage unit 8 is "01", the present embodiment is step S1107.
Then, after assigning 1 to turn, the process proceeds to step S1109. The 2-bit data received from the first storage unit 8 is "1".
In the case of 0 ', in this embodiment, in step S1108, turn
After substituting 2 for, the flow advances to step S1109. That is, the 2-bit data received from the first storage unit 8 is "0".
In the case of 0 ',' 01 ', and' 10 ', the output pixel data are pixel data latched by former [0], former [1], former [2], respectively.

Subsequently, in this embodiment, step S1109 is executed.
Then, after replacing the pixel data latched in next with the pixel data latched in former [turn], step S1
Proceed to 113. On the other hand, if the 2-bit data received from the first storage unit 8 is "11", the present embodiment is directed to step S
At 1110, a latch data number different from turn_p and turn_pp is set to turn, and at step S1111 the former [tur
Substituting the pixel data latched in next into [n], and in step S1112, after substituting the 24-bit pixel data received from the first storage unit 8 into next, the process proceeds to step S1113. Note that in steps S1103 and 1104 described above,
Since turn_p and turn_pp are not set to the same value, turn is 0, 1, or 2 in step S1110.

Subsequently, in the present embodiment, step S1113 is executed.
Output the pixel data latched to next as a decompression result, increment the counter i in step S1114, and determine the value of the counter i in step S1115,
If i ≦ 63, the process returns to step S1102 to process the remaining pixels, and if i> 63, the process returns to the processing procedure of FIG.

As described above, according to this embodiment,
In order to prevent image deterioration and achieve the target compression ratio, some image data can be losslessly compressed, and other image data can be lossy compressed, and the compressed data can be appropriately expanded according to the compression method. Furthermore, according to the present embodiment, lossless compression data and lossy compression data can be decompressed in block units. Therefore, for example, even for a natural image embedded in a computer created image, a computer created image embedded in a natural image Also works effectively against.

[0090]

[Fourth Embodiment] In the above-described embodiment, the memory is divided into segments in order to achieve the target compression ratio, and the segment controller is used. However, in the first compression process, the area information is stored in the second storage unit. May be generated, the delta factor of ADCT that achieves the target compression rate may be calculated, and this may be achieved by the second compression processing. In this case, complicated memory control processing becomes unnecessary.

[0091]

[Fifth Embodiment] In the first and second embodiments described above, the reversible compression data output from the reversible compression unit 3, that is, the signals S2a and S2b are turned ON / OFF ('1').
/ '0') is configured by the flag system,
In the fourth embodiment, the signal is constructed by a trigger system.

13A and 13B are views for explaining the difference between the flag system and the trigger system. FIG. 13A shows the flag system and FIG. 13B shows the trigger system. As is clear from FIG. 13, the same data tends to last longer in the trigger system. This embodiment utilizes this property to further add M to the signals S2a and S2b.
A run-length compression such as H is performed to reduce the amount of data.

Further, not limited to MH, Huffman coding, Lempel-Ziv coding, etc. may be applied to the completed compressed data to further compress the data. The signal S
It is also effective to apply Huffman coding or Lempel-Ziv coding to 2c. The present invention may be applied to a system including a plurality of devices or an apparatus including a single device.

It is needless to say that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0095]

As described above, according to the present invention, the relation between the compressed data amount D and the predetermined value L when the block is losslessly compressed is
If D ≦ L, the block is losslessly compressed and D>
If it becomes L, either the result of lossy compression processing of the block and decompression processing of the compressed data as lossless compression data or the result of decompression processing of the compressed data as lossy compression data is input information. It is possible to provide an image processing method and an image processing apparatus that restore the original image data by selecting based on the.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration example of an image compression apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration example of a lossless compression unit according to the present exemplary embodiment.

FIG. 3 is a block diagram showing a configuration example of an area determination unit of this embodiment.

FIG. 4 is a block diagram showing a configuration example of an ADCT unit of this embodiment.

FIG. 5 is a flow chart showing an example of a reversible compression procedure according to the second embodiment of the present invention.

FIG. 6 is a flowchart showing an example of the processing procedure of step S404 in FIG.

FIG. 7 is a flowchart showing an example of the processing procedure of step S506 in FIG.

FIG. 8 is a diagram showing an example of the processing order of block pixels according to the present embodiment.

FIG. 9 is a block diagram showing a configuration example of a data decompression device according to a third embodiment of the present invention.

FIG. 10 is a flow chart showing an example of a processing procedure for realizing the lossless compression data decompression unit of the present embodiment by software.

11 is a flow chart showing the procedure of extension processing in step S1003 of FIG.

FIG. 12 is a block diagram showing another configuration example of the reversible compression unit of the first embodiment.

FIG. 13 is a diagram illustrating a difference between a general flag system and a trigger system.

FIG. 14 is a block diagram showing a configuration of a conventional ADCT image compression device.

FIG. 15 is a block diagram showing a configuration of a conventional ADCT data decompression device.

FIG. 16 is a diagram showing a part of image data of a general color multi-valued image created by a computer.

17 is a diagram illustrating a conventional ADCT based on the image data shown in FIG.
It is a figure which shows the result processed by the data compression apparatus.

FIG. 18 is a diagram showing a state in which 8 blocks shown in FIG. 17 are DCT-transformed.

FIG. 19 is a diagram showing a general quantization table used for quantization.

20 is a diagram showing a result of quantizing the DCT coefficient data shown in FIG. 18 using the quantization table shown in FIG.

FIG. 21 is a diagram showing a general state of rearranging pixel data in zigzag order.

FIG. 22 is a diagram showing the result of Huffman decoding and inverse quantization performed on general ADCT compressed image data.

23 is an inverse DCT of the DCT coefficient data shown in FIG.
It is a figure which shows the result.

FIG. 24 is a diagram showing data finally obtained by a conventional ADCT data decompression device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image data storage unit 2 Delay unit 3 Reversible compression unit 4 ADCT compression unit 5 Buffer 6 Segment controller 7 Segment information table unit 8 First storage unit 9 Second storage unit 10 Control unit 201 Rearrangement unit 202 to 205 First to first 4 latch unit 206-209 1st-4th comparison unit 210 decoder 211 PLD 213 back-up unit 214 area determination unit 801 color space conversion unit 802 DCT unit 803 quantization unit 804 encoding unit 901 segment controller 902 ADCT decompression unit 904 lossless compressed data Decompression unit 906 Selector 907 Image data storage unit 910 Control unit

Claims (9)

[Claims] 1. An image processing method for processing image data in blocks, wherein the block is reversibly compressed when the relationship between a compressed data amount D and a predetermined value L is D ≦ L. An image processing method, wherein lossless compression processing is performed, and when the relationship between the compressed data amount D and the predetermined value L when the block is losslessly compressed is D> L, the block is lossy compressed. 2. The reversible compression process includes a rearrangement step of rearranging the pixel data of the block one-dimensionally and outputting the pixel data, a latching step of latching the pixel data output in the rearrangement step, and the latching step. A first pixel comparison step of comparing the first pixel data latched by the first pixel data and the second pixel data output in the rearrangement step subsequent to the first pixel data, and the first pixel comparison A holding step of holding at least one pixel data latched by the latch means when the comparison result of the process does not match, and the holding process when the comparison result of the first pixel comparison process does not match. The second pixel comparison step of comparing the pixel data held in the step with the second pixel data, and the second pixel comparison step when the comparison result of the first pixel comparison step does not match. comparison In accordance with the result, one of the pixel data held in the holding process is updated to the pixel data latched in the latching process, and when the comparison result of the comparing process is D> L, the latching process is performed. A recovery step of returning the latched pixel data and the pixel data held in the holding step to a value before the processing of the current block, the comparison result of the first pixel comparison step and the second pixel comparison step. Is output, and if the second pixel data does not match any of the pixel data held in the holding process by the second pixel comparison process, the second pixel data is output. The image processing method according to claim 1, wherein the image is output. 3. An image processing apparatus for processing image data in block units, comprising: first compression means for reversibly compressing the block; second compression means for irreversibly compressing the block; A comparison means for comparing the compressed data amount D and a predetermined value L when the block is compressed by the compression means, and a compression result of the first compression means if the comparison result of the comparison means is D ≦ L. And selecting means for selecting the compression result of the second compression means when the comparison result of the comparison means is D> L, the comparison result of the comparison means is selected by the selection means. An image processing apparatus which outputs the compressed result. 4. The image processing apparatus according to claim 3, wherein the second compression means performs compression by an adaptive discrete cosine transform method. 5. The first compressing means rearranges means for rearranging the pixel data of the block one-dimensionally and outputs the pixel data, and a latch means for latching the pixel data output from the rearrangement means. First pixel comparison means for comparing the first pixel data latched by the latch means with the second pixel data output from the rearrangement means subsequently to the first pixel data; and the first pixel comparison means. Holding means for holding at least one pixel data latched in the latch means when the comparison result of the pixel comparison means is inconsistent, and the comparison result of the first pixel comparison means is inconsistent Second pixel comparing means for comparing the pixel data held in the holding means with the second pixel data; and the second pixel if the comparison result of the first pixel comparing means does not match. Comparison means Updating means for updating one of the pixel data held in the holding means to the pixel data latched by the latch means according to the comparison result, and when the comparison result of the comparing means is D> L, the latch Means for recovering the pixel data latched by the means and the pixel data held by the holding means to a value before the processing of the current block, the comparison result of the first pixel comparing means and the second pixel The second pixel is output when the comparison result of the comparison means is output, and when the second pixel comparison means does not match any of the pixel data held in the holding means by the second pixel comparison means. The image processing apparatus according to claim 3, wherein the image processing apparatus outputs data. 6. The image processing apparatus according to claim 5, wherein the rearrangement unit scans the blocks in a zigzag manner. 7. The comparing means counts the number of times that the comparison result of the first pixel comparing means becomes inconsistent, and the second pixel comparing means counts the second count. Second counting means for counting the number of times that the pixel data does not match any of the pixel data held in the holding means, the coefficient result of the first counting means and the counting of the second counting means. The image processing apparatus according to claim 5, wherein the compressed data amount D is calculated from the result. 8. The comparing means includes an accumulating means for accumulating a difference S between the predetermined value L and the compressed data amount D of the processed block, and when the comparison result in the current processing block is D ≦ L, the predetermined value is the predetermined value. 8. The image processing apparatus according to claim 3, wherein the value L is updated to a sum of a predetermined value C and the difference S accumulated in the accumulating unit. 9. An image image processing method for decompressing compressed data obtained by compressing image data in block units, comprising: a result of decompressing the compressed data as lossless compressed data and decompressing the compressed data as lossy compressed data. An image processing method, characterized in that any one of the results is selected based on the input information, and the selected expansion results are combined to restore the original image data.