JPS5814674A - Image coding system - Google Patents

Abstract

PURPOSE:To make an initial block size most suitable to every object picture element, by dividing a subject image into fine blocks, repeating more divisions only to the block containing the picture elements other than the white ground and coding the gradation information for only the block that has a 2X2 picture element size. CONSTITUTION:An original image is set at an NXN picture element size, and the gradation information per picture element is set at (a) bit. This image is divided into the blocks of 2<k>X2<k> picture element size, and the coding is performed with each bit per block to decide whether each block has the white ground or contains the picture elements (black ones) other than the white ground. Then the block containing the black picture elements is divided into minor blocks of 2<k-1>X2<k-1> picture element size. Then each minor block is coded with every bit to decide whether each minor block has the white ground or contains the black picture elements. Furthermore the minor block containing black picture elements is divided into half, and finally the gradation is coded with each picture element by the (a) bit for the block of 2X2 picture elements containing black picture elements.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an image encoding method for efficiently encoding a grayscale image containing a large amount of white background with a small amount of code.

Conventionally, various encoding methods such as run-length encoding and READ encoding have been developed as so-called facsimile signal encoding methods for binary images containing many white backgrounds. Coding methods such as block coding methods have been developed for general grayscale images that do not have a gradation image.

However, conventional facsimile signal encoding methods cannot encode images in which a pattern with some shading, such as a seal imprint or a fingerprint, is drawn on a white background.

In addition, it is possible to apply a grayscale single image encoding method such as a block encoding method to such an image as a type of grayscale image, but it is not very redundant because the condition that there are many white backgrounds is not actively used. The drawback was that the effect of suppression could not be expected.

In order to solve these drawbacks, the present invention divides the target image into small blocks, and then divides the blocks further only for blocks containing pixels other than white, and finally reduces the size to 2K2 pixels. This method applies a method that encodes grayscale information only for blocks, and in particular optimizes the block size at which division begins for each target image to improve encoding efficiency.The drawings are explained in detail below. .

FIG. 1 illustrates the principle of the present invention. As shown in the figure, the original image has a size of N×N pixels, and the grayscale information per pixel is α pit. This is first divided into blocks each having a pixel size of 2×2, and whether each block is a white background or contains pixels other than a white background is encoded using one bit per block. Next, for blocks containing pixels other than white background, add 2 more
Divide -xX2- into small blocks of pixel size,
Whether each small block is white or contains pixels other than white is encoded using one bit for each small block. Furthermore, the small block containing pixels other than the white background is subdivided into halves, and the same operation is repeated until the block is divided into 2×2 pixel blocks, thereby performing multi-stage division. Finally, for the 2K2 pixel block containing pixels other than the white background, the shading is encoded for each pixel using α pits, that is, 4α bits per block.

In the above, when the original image of NXN pixels is divided into 2 x 2 blocks of pixel size, the number of blocks is n(2
k) (2N2/4k), and the number of blocks containing pixels other than white is rL (2k), then the original image is divided into 2k x 2 blocks starting from blocks of pixel size in multiple stages and coded. The code amount C(2) when converting is C(2k) ``4o(2k) +4 W (2k)+
43 (2 to 1') +−−−・−・+4 =8)+4
It is expressed as 7+(4)+412m2)...(]). Here, when the number of multi-stage divisions of the original image is less than 1 step, that is, when the initial block size is 61× and 1 pixel, the code amount C(2k
) is the increment of ΔC(2k) (k>2), then rL
Since o(2 to -1)=4rLo(2k), using equation (1), 110(2")=4W(2k)-9,(2k)
C&>1) (2). Then, ΔC(2'') - 4rL (2 to 1') - 3Wo (2 to 1')
) = 4 (W (2 to -1) -4jL (2k) + ΔC (2
k)) (3). Here, always 7L (2”)≦4r
Since L(2k), ΔC(2"), 44ΔC(2'
) (4) Totoru. Therefore ΔC(
2 &) If <0, it will always be -C(2'-1K O.This means that the maximum integer value k such that ΔCC2S<O gives the optimal initial block size when dividing the original image into multiple stages. ( In formula 2), the initial number of divided blocks is 71O.
The number of white blocks among (2k) blocks is yv
,, (2k) pieces (~<2s=no<2S<2S), ΔC(2”)=4°(2k)−4rLW(2k)(5)
becomes. Therefore, from equation (5), when a given image is divided into multiple stages and encoded, the most coded image can be obtained by dividing from the largest block size such that blocks larger than the total number of blocks V4 are blank. It is concluded that the conversion efficiency is good.

Note that the difference in code amount ΔC(2
) is from equation (1), ΔC(2)=C(2)−4αno
(2) = 4α bite) - 4αW, (2) + W. (2
), and the cardinal point 2)-ru. (2) −rL, substituting (2) gives ΔC(2) = W. (2) -41? L, (2
) (6) becomes. Therefore, as mentioned above, nvp) > 4yLo (Even if it is not 2, if the number of white background blocks occupies a larger proportion than 4a to the total number of blocks when the original image is divided into blocks of 2 x 2 pixels size, then 2 An encoding method that divides into blocks of ×2 pixel size is effective for reducing the amount of code.

FIG. 2 shows an example of a configuration for realizing this encoding method. In the figure, 1 is a frame memory, 2 is a window circuit, 3 is a register, 4 is a black pixel detector, 5 is a register selection circuit, 6-1.6-2. ......, 6- is a register, 7-1.7-2.・・・・・・・・・7
- is a register, 8 is a register selection circuit, 9 is a comparator,
10 is a register selection circuit, 11-1.11-2. ...
・Song・, 11- is a register, 12 is a register, and 13 is a code string reading circuit.

To explain this operation, first, an original image of N×N pixels stored in the frame memory 1 is sequentially cut out by the window circuit 2 into blocks each having a size of 2k×2 pixels.

Here, k is a predetermined initial value, and a value of 2 is stored in the register 3 as the cutout block size value of the window circuit 2. A black pixel detector 4 detects whether each block of the original image cut out by the window circuit 2 contains pixels other than white pixels (referred to as black pixels), and detects whether each block contains, for example, black pixels. The case of a white background is encoded with the bit "OI".The code string obtained by the black pixel detector 4 is sent to the register selection circuit 5, and the bit "in the code string"0'', that is, the number of white blocks ~(2k), and store the code string and the number of white blocks in registers 6-k and 6-k, respectively.
Store in register 7-k. The number of white blocks W (2k) stored in the register 7-k is sent to the comparator 9 through the register selection circuit 8, and compared with the contents of the register 1- selected by the register selection circuit 10. Here, register 1-k contains a value obtained by dividing the total number of blocks rLo (2k) by 4 when the original image previously stored in the frame memory is divided into blocks each having a pixel size of 2×2. (2k)/4 is stored as a known value. The comparator 9 is the above rLW(2k). (2k)/4 is compared, and when Flw(2k)≦no(2k)/4, an output "No" is output, and when FLw')>n (2")/4, an output "-YES" is output. If comparator 9 outputs 1N01, it is assumed that the initial block size when dividing the original image into multiple stages is too large, and the contents 2k of register 3 are shifted by 1 bit in the low bit direction and changed to 2 by -1. , the block size from which the window circuit 2 cuts out the original image is set to 2 -I x 2 & -1 pixels. Also, the comparator 9 simultaneously selects the register selection circuits 5, 8, and 1.
0, each register 6-(k-1), the same?
-Ck-1) and 11-C' are made to select -1). Register 1-(k-1) contains the total number n of blocks when the original image is divided into 2:1 x 2:11 pixel blocks. (2”) divided by 4 n. (2”)
/4 is stored. Using the same procedure as above, the window circuit 2 divides the original image into blocks of size 2kI x z&-1 pixels and stores the number of white blocks W (2 to 1') in the register 7-(&-1). The value n stored in sister 1-(k-1). (2 to 1'V4
are compared by the comparator 9. If rL (2 to 1')
4n (2 to 1')/4, comparator 9 indicates I No
Output I and repeat the above operation. By repeating the above operations, we will tweet at the stage where nw(2″′)〉no(2rn)74μrum is obtained.WW(2″)
')〉When no(2rn)/4, comparator 9 is 'YES'
The output is input to the register selection circuit 5 as an operation command signal. Upon receiving the operation command signal, the register selection circuit divides the code string stored in the register 6-m, that is, the original image, into blocks each having a size of 2rn×2''!pixels, and determines whether each block contains black pixels or not. Reads out a code string indicating whether the
``The output is input to the register 3 at the same time as the register selection circuit 5, and the register 3 shifts its contents, that is, the value 2m, which gives the divided block size of the original image, by 1 bit in the low pit direction and converts it into the value 2m-1. The window circuit 2 receives the contents of the register low m from the register selection circuit 5, and converts the original image only to the 2m x 2 pixel size block including black pixels into the 2m contents of the register 3.
Referring to the value -1, the block is sequentially divided into small blocks of 2m-1×2771-1 pixels. divided 2″
The black pixel detector 4 detects whether or not a black pixel is included for each block having a size of L-1×2771-1 pixels, and the result is encoded with I11 if it includes a black pixel and 04 if it is white. do. The register selection circuit 5 stores the code string in the register 6-(m-1), and at the same time operates the window circuit 2 again using the code string to select 2m-1X including black pixels.
For only blocks with a size of 2WL-1 pixels, the original image is further divided into blocks with a size of 2m-2 x 2''L-2 pixels.In each case, the register selection circuit simultaneously selects the contents fL-' of register 3. It is shifted by 1 bit in the low pit direction and converted to a value of 2. This value is referred to as the divided block size value of window circuit 2. 2
Repeat until it is divided into blocks of size ×2 pixels. When the division into blocks with a size of 2*2 pixels is completed, the register selection circuit 5 operates on the window circuit 2 and selects the original image pixel by pixel only for blocks containing black pixels among the blocks with a size of 2*2 pixels. To divide. At the same time, the register selection circuit 5 shifts the content 2 of the register 3 by 1 bit in the low pit direction and converts it to a value 1, and this value defines the divided block size of the window circuit 2. The pixel data string divided by the window circuit 2 does not pass through the black pixel detector 4, but is stored in the register 12 through the register selection circuit 5 while retaining grayscale information for each pixel. When the above operations are completed, the code string reading circuit 13 reads the registers 5-m, 6-(m-1), 4..., 6 through the register selection circuit 5.
-1, the contents of the register 12 are read out as an image code string, and the number of read registers (m+1) is simultaneously included in the image code string as the number of division stages of multi-stage division.

Note that in the above operation, register 11'(' 2 * 3
1...-..., &) is the total number of broncs when the original image is divided into blocks of size zt x 2Z pixels.

The value n obtained by dividing (24) by 4. (2t)/4 is stored, and the comparator 9 compares the size with the number of white blocks n (2t) in the same division.As shown in the above principle, the contents of the register 11-1 are Le. (2)/4
11 ((L is the number of bits to represent the shading per pixel of the original image) is stored, and t′Lw(2)〉n
If o(2)/4α, only one stage of divisional encoding will be performed.

As explained above, by using the present invention, it is possible to encode a grayscale image containing a large amount of white background with a small amount of code, so it is possible to reduce the file size when a large number of files of the above-mentioned images, such as seal impressions and fingerprints, are stored. , and when transmitting the images at the same time, there is an advantage that the images can be transmitted using a transmission path with a small transmission capacity.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the principle of the present invention, and FIG. 2 is an example of a configuration method for realizing the system of the present invention. 1......Frame memory, 2...
・・・・・・Window circuit, 3 ・・・－・・・・・・
・Register, 4... Black pixel detector,
5......Register selection circuit, 6-1.
6-2゜・・・・・・・・・、(、-k・・・・・・・・・
...Register, 7-1.7-2.・・・・・・・・・、
7-k・・・・・・Register, 8・・・・・・
...Register selection circuit, 9...Comparator, 10...Register selection circuit, 11
-1.11-2. ......, 11-/c...
......Register, 12...Register, 13... Code string readout circuit.

Claims (2)

[Claims] (1) A means for storing a grayscale image to be encoded, a means for dividing the grayscale image into a plurality of blocks with a specified block size, a means for examining the grayscale distribution state of each divided block, and expressing it as a code string. means for checking the number of blocks that can be represented by one gray level; means for storing the checked gray level distribution state and the number of blocks; and the total number of divided blocks and the blocks that can be represented by the one gray level. means for detecting whether the currently specified block size is the optimal initial value block size by comparing the An image encoding method characterized in that only blocks whose gradation distribution state cannot be represented by one gradation level are subjected to divisional coding by repeatedly dividing them into blocks of smaller size. (2) When the original image is divided into blocks, the maximum block size such that more blocks than the total number of blocks become white is the optimum initial value.