JPH0614738B2 - Image compression method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression method, and more particularly to an image compression method suitable for reducing noise in a reproduced image.

[Background of the Invention]

When using an image having an enormous amount of information as an information medium, a technique for compressing image information has become indispensable.

As a conventional technique for compressing image information, for example, the Institute of Electronics and Communication Engineers, '85 / 9VoL. J88-BNO.9, 1088 describes a method of Hadamard-converting an original image to obtain converted data, and changing the number of bits to be distributed to each converted data according to the characteristics of the converted data.

To increase the compression rate, the number of allocated bits may be reduced.
However, if the number of allocated bits is reduced, noise will increase in the reproduced image and the difference from the original image will increase. Therefore, it is general to increase the number of allocated bits only at a specific portion of the image. Although the quality of the reproduced image is determined by where many bits are allocated, the above-mentioned conventional image compression method allocates bits using only the characteristics of Hadamard-converted data. When doing so, there is a problem that the reproduced image contains a lot of noise and the visual characteristics are not sufficient.

[Object of the Invention]

An object of the present invention is to provide an image compression method in which noise in reproduced images is small and visual characteristics are good.

[Outline of Invention]

To achieve the above object, in the present invention, a difference image between an original image and an image obtained by enlarging a reduced image of the original image is orthogonally transformed, and the orthogonal transformation data is digitized by referring to a bit allocation table. In the lossless image compression method, the orthogonal transformation data of the difference image that has been digit-carried by referring to the bit allocation table is carried by referring to the bit allocation table.
The difference data between the orthogonal transform data before the carry and the carry after carry is checked to correct the bit allocation of the bit allocation table, and the orthogonal conversion data of the difference image is subjected to the digit compression by using the modified bit allocation table and compressed. .

Example of Invention

 An embodiment of the present invention will be described below with reference to the drawings.

FIG. 6 is a hardware configuration diagram for carrying out an image compression method according to an embodiment of the present invention, which includes a central processing unit (CPU) 1, a high-speed arithmetic circuit 2, a storage medium 3, a display 4, and an image (not shown). The reading device is connected to each other by a bus 5. The CPU 1 controls / accesses the high-speed arithmetic circuit 2 and the storage medium 3 to instruct the high-speed arithmetic circuit 2 to perform orthogonal transformation, which will be described later, and store data in the storage medium 3.

FIG. 1 is an explanatory diagram of an image compression method according to an embodiment of the present invention. In this embodiment, as the first step, the original image
(a) 6 is reduced to obtain a reduced image (S) 7, and this reduced image (S)
7 is stored in the storage medium 3. Further, the reduced image (S) 7 is interpolated and enlarged to make an enlarged image (a ′) 8.

As a second step, a difference image (A = a−a ′) 9 between the original image (a) 6 and the enlarged image (a ′) 8 is created, and using this and a bit allocation table (BAT) 10, bit allocation is performed. Table correction process
11 is performed to create a modified bit allocation table (BAT ').

Details of the bit allocation table correction processing 11 will be described with reference to FIG. The bit allocation table correction process is roughly divided into stage I and stage II. In stage I, first, the image (A
= A-a ') 9 is subjected to orthogonal transformation such as Madamale transformation to obtain transformed data (B) 13. Next, referring to the bit allocation table 10, the converted data 13 is digit-decimated and the digit-converted data is digitized.
Make 14 The bit allocation table 10 may be appropriately determined initially. For example, a histogram of data is obtained and the distribution bits are uniformly set to 4 bits. Then, the converted data 13 is converted to 16
Shift 12 bits out of the bits to reduce the value, and generate the digit conversion data 14. Then bit allocation table 10 again
The digit conversion data 14 is reversed by 12 bits to carry a digit to obtain converted data (B ′) 15. This converted data 1
Image (A) 9 reproduced image (A ') 16
make. Converted data (B ') 15 is digitized from converted data (B) 13 with reference to the roughly made bit allocation table 10.
Since it was created by carrying, B ≠ B. Therefore, image A ≠ reproduced image A ′.

In step II, a difference image (AA ') 17 is created by subtracting the reproduced image A'from the image A, and this is orthogonally transformed to obtain the transformed data 1
Ask for 8. The converted data 18 is examined to find a portion having a larger value than the others. Assuming that the larger portion than the other is x, y, the distribution bit number of the corresponding portion x, y of the bit distribution table 10 is increased, and the modified bit distribution table 12 is obtained.

Modified Bit Allocation Table 12 modified in this way
Is more accurate than the original bit allocation table 10. If a more accurate modified bit allocation table is desired, it can be obtained by repeating the procedure of steps I and II with the modified bit allocation table 12 as the bit allocation table 10 of step I.

In the bit allocation table correction process shown in FIG.
The difference image (AA ') between (A) and its reproduced image (A') is orthogonally transformed to find the bit allocation correction points x and y. As shown in FIG. The difference 18 'between the converted data (B) 13 and the converted data (B') 15 in I is checked to find the bit allocation correction points x and y, but they are mathematically equivalent.

After the modified bit allocation table 12 is created as described above, the difference image (A = aa ') 9 is compressed using the modified bit allocation table 12 in the third stage of FIG. First, the difference image 9
Is orthogonally transformed to create transformed data 19. Then, with reference to the modified bit allocation table 12, digit cancellation is performed to generate the digit conversion data 20, and this is filled as described later to obtain the filling data (CP) 21 in order to reduce the capacity, and the storage medium 3 is obtained.
To store. Further, the modified bit allocation table 12 is also filled to obtain the filled bit allocation table 22 and stored in the recording medium 3.

"Filling" refers to the following processing. For example, when 16 bits are allocated to each pixel of the image and the image is converted into a digital signal and stored in the memory, storing one pixel at each address of the memory configured in 16 bits requires a huge memory capacity. Therefore, when a variable-length code such as Huffman code is used, and the pixel information with a higher appearance probability is assigned a shorter code length (bit number), the image is digitized and stored in the memory. Store them together without any gaps. This is called filling. That is, pixel data represented by a code length different from the 4-bit configuration, 5-bit configuration, and 7-bit configuration are sequentially stored in a 16-bit unit memory without any space. As a result, for example, as shown in FIG. 4, when 6 pixels each represented by 16 bits are stored in the memory, a capacity of 6 addresses × 16 bits is required. × 16-bit capacity will be enough. When the image data is read from the filled and stored memory, the original data before the filling can be restored by referring to the table storing the rule of the variable length code such as Huffman code. This is called "extension".

Next, the procedure for reproducing the image a will be described with reference to FIG. From the storage medium 3, the reduced image (S) 7 and filling data (C
P) 21 and the filling bit allocation table 22 are read. The reduced image 7 is interpolated and enlarged to form an enlarged image (a ′) 8. The filling data 21 and the filling bit allocation table 22 are returned to the digit conversion data 20 and the correction bit allocation table 12 by performing the above-mentioned "expansion" processing, respectively. And the digit conversion data 20
The carry is carried out while referring to the modified bit allocation table 12, and the carry conversion data 23 is created. When this carry-converted data 23 is subjected to inverse orthogonal transform, a reproduced image 24 of the difference image (aa ') is obtained.
By adding the reproduced difference image (aa ') 24 and the enlarged image (a') 8, a reproduced image 25 of the image a is obtained.

The reproduced image 25 obtained as described above reproduces the original image 6 (FIG. 1) with high accuracy and has little noise and excellent visual characteristics.

〔The invention's effect〕

According to the present invention, since the noise in the reproduced image is small and the difference from the original image is small, the visual characteristics and fidelity of the reproduced image are improved.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of an image compression method according to an embodiment of the present invention, FIG. 2 is a detailed explanatory diagram of the bit allocation table correction processing shown in FIG. 1, and FIG. 3 is another procedure of the bit allocation table correction processing. 4 is an explanatory view of the filling method shown in FIG. 1, FIG. 5 is an explanatory view of the image reproducing method, and FIG. 6 is a hardware configuration diagram. 3 ... Storage medium, 6 ... Original image, 7 ... Reduced image, 8 ... Enlarged image, 9 ... Difference image, 10 ... Bit allocation table, 12 ... Modified bit allocation table, 24 ... Reproduction difference image, 25 ... Reproduction image.

Claims (1)

[Claims] 1. A irreversible image compression method for orthogonally transforming a difference image between an original image and an image obtained by enlarging a reduced image of the original image, and carrying out digit shift of the orthogonal transformation data while referring to a bit allocation table, The orthogonal transformation data of the difference image, which has been digit-shifted by referring to the bit allocation table, is carried by the digit by referring to the bit-allocation table, and the difference data between the orthogonal transformation data before and after the digit cancellation is checked to determine the bit. An image compression method, characterized in that a correction process for correcting the bit distribution of a distribution table is performed and the orthogonal transformation data of the difference image is compressed using the obtained modified bit distribution table.