JPH0426277A - Image processing method - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Summary] Regarding an image processing method when obtaining a pseudo halftone image by binarizing density data, the present invention relates to an image processing method for obtaining a pseudo halftone image by binarizing density data. For the density data, perform Laplacian calculation between the pixel of interest and its surrounding pixels, calculate the difference between the fixed value determined by the number of bits of the density data and the Laplacian calculation result, and divide the difference calculation result into 2 Binarize the pixel of interest as a threshold value for value conversion, and
A difference calculation is performed between the density data of one pixel and the data obtained as a result of binarization, and the result of the difference calculation is used as an error, and an error diffusion calculation is performed on surrounding pixels using an error diffusion matrix to calculate the error. Configure to make allocations.

[Industrial Field of Application] The present invention relates to an image processing method for obtaining a pseudo halftone image by binarizing density data.

In devices that perform digital image processing, such as digital copying machines, image scanners, facsimile machines, etc., halftone image processing is performed to obtain pseudo halftone images (expressed using binary values of 0 and 1).

[Prior Art] As a halftone image processing method in this type of apparatus, the dither method has been widely and generally used. FIG. 5 is an explanatory diagram of the DISA method. For the original picture shown in (a), (
When the image is binarized using a threshold value as shown in b) (the portion surrounded by a thick solid line in the 4×4 dither matrix 1), a binarized image as shown in (C) is obtained. Here, the dither matrix in (b) is repeated in units of 4×4 pixels.・(
“1”) is obtained when pixel density ≧ threshold value, ○ (“0”)
is obtained when the pixel density is less than the threshold value. In (a), one square or one pixel is represented.

The dither method as described above has the advantage of being able to express pseudo halftones at low cost because the hardware configuration is simple. On the other hand, (1) When the original is a halftone image such as a printed image, periodic striped patterns (moiré) that are not present in the original occur in the processed image. The cause of this is shown in Figure 5(b).
It is based on having a dasa matrix or periodicity as shown in .

■If the original contains text, line drawings, etc., the processing results for those parts will be cut off and reproducibility will be extremely poor.

■If the matrix size is made too large relative to the resolution of the reading system in order to obtain multiple gradations, the resolution will deteriorate, making it impossible to achieve both multiple gradations and high resolution.

There is a problem.

On the other hand, in order to solve this problem, there is a method called the error diffusion method, which is a halftone image processing method that can achieve both multi-gradation and high resolution. This method is
R,W,Floyd and L,Steinbe
rg"An Adaptive Algorithm
m for 5special Gray 5c
ale”1975 SID International
nal Symposium Digest o
f Techrlica I Papers, 4.3
．． It was published in pp 36-37 (Apr, 1975).

FIG. 6 is an explanatory diagram of the error diffusion method. Now, there are density data D1.1 to D33 as shown in FIG.
22 (indicated by diagonal lines) is the pixel of interest.

It is assumed that these data Dll to D33 are, for example, 8-bit data. In (b), the density value D22 of the pixel of interest is 2.
25. At this time, the full scale value is 25
The difference from 5 is -30.

On the other hand, as shown in (C), pixels D23 to D33 that come after the pixel of interest (indicated by diagonal lines) are weighted as shown in the figure. In other words, the pixel D2 closest to the pixel of interest
A weight of 2 is assigned to pixels D3 and D32, and a weight of 1 is assigned to the other pixels D31 and D33.

The reason why the pixels D11 to D21 before the pixel of interest D22 are not taken into consideration is because these pixels have already been binarized.

The above-mentioned difference -30 is regarded as an error, and this error is assigned to the pixels D23 to D33 around the pixel of interest D22 in the following manner.

D23・-30X2/6--ICI D31 ;-30xl/6--5 D32 Knee 30X2/6==-] 0D33, -30
X1/6--5 As a result, the density value D23-- of pixels D23 to D33
D33' are as follows.

D2B-=D23-10 D31--D31-5 D32--D32-10 D33-=D3B-5 The density value thus determined is determined by a predetermined threshold value (intermediate density value, for example 127 in the case of 8-bit data). Binarize.

As explained above, the error diffusion method is a method characterized by diffusing the density error that occurs during binarization to surrounding pixels so that the density of the original image and output image is preserved. Unlike the dither method, the expressed gradation is not determined uniquely by the matrix size, and moiré does not occur, making it an excellent method in terms of gradation and single resolution.

[Problems to be Solved by the Invention] However, since 2, the error diffusion method also has the following problems.

■Compared to simple binarization processing, the reproducibility of characters and line drawings is also inferior.

■In areas with low density, errors gradually accumulate and then
”, there is a spatial delay in the appearance of the dot.

The present invention has been made in view of such problems, and an object of the present invention is to provide an image processing method that can obtain good pseudo-halftone images for both characters, line drawing parts, photographs, and halftone parts. It is said that

[Means for Solving the Problems] FIG. 1 is a flowchart showing the principle of the method of the present invention. The present invention performs a Laplacian calculation between the pixel of interest and its surrounding pixels on the density data stored in the density data line buffer (step 1), and calculates a fixed value determined by the number of bits of the density data and the Laplacian. A difference calculation is performed with the calculation result, and the pixel of interest is binarized using the difference calculation result as a threshold for binarization (step 2), and the density data of the pixel of interest and the data obtained as a result of the binarization are A difference calculation is performed (step 3), and the result of the difference calculation is used as an error, and an error diffusion calculation is performed on surrounding pixels using an error diffusion matrix to perform error distribution (step 4). It is said that

[Operation] A Laplacian calculation is performed on the pixel density data between the pixel of interest and its surrounding pixels, a difference calculation is performed between a fixed value determined by the number of bits of the density data and the Laplacian calculation result, and the difference calculation result is calculated. is used as the threshold for binarization. The Laplacian operation is obtained as the sum of the differences between the densities of the pixel of interest and surrounding pixels on a discrete space, and the Laplacian operation result represents the amount of edge, and its absolute value increases in the contour part of a single line drawing of a character. In other words, by changing the threshold for binarization based on this value, the threshold for binarization becomes smaller for black parts, and the threshold for binarization becomes larger for white parts. contours) are emphasized, and a good pseudo-halftone image can be obtained.

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 2 is a diagram showing an example of a Laplacian filter (coefficient matrix) that performs a Laplacian operation, showing three types (a) to (C). In the discrete space, the Laplacian operation is calculated as the sum of the differences in density between the pixel of interest and surrounding pixels, as described above.

In either case, the pixel is the center pixel or the pixel of interest. For example, (a)
For example, in the example shown in the figure, only the pixels to the top, bottom, left, and right of the 1st pixel are considered, and when performing the Laplacian operation, the pixel of interest is added four times, and the pixels to the top, bottom, left, and right are subtracted once. It is shown that.

For example, assuming that there is density data as shown in Figure 3, if Figure 2 (a) is applied to this density data,
The difference calculation is (30-10) + (30-20) + (3(L-8)
+ (30-8)=74. In reality, 2/74 is too large as a difference result, so processing such as multiplying it by a certain coefficient K (for example, 0.5) is performed.

FIG. 4 is a diagram showing an example of a circuit configuration for implementing the present invention.

In this embodiment, the binarization threshold is corrected by Laplacian calculation, and the binarization process is performed by the error diffusion method. First, an image is read by a line image sensor in which N (N is an integer) dots are lined up in the main scanning direction from an image reading unit (not shown in the figure), and the original image data is quantized according to the density (i: small density: 0, maximum density: Density 255) is input. This original image data is stored in the corresponding position of the density data line buffer 1. In the figure, the density data line buffer 1
indicates a 3×3 matrix part.

The density data line buffer 1 is composed of, for example, a RAM, and stores density data for the pixel of interest to be processed now and a total of three lines before and after it.

In other words, the column direction is the (n-1)th dot, the nth dot, the so-so (n+1)th dot, and the row direction is the (ml)th dot.
There are three lines: line 1, mth line, and (m+1)th line. These pixels are sequentially D□-1, n-1 to [)z+1.
n"l.The pixel of interest is D4. of the n-th dot of the m-th line. Note that m and n are positive integers, and 1≦
Let n≦N.

The density data of the pixel of interest and its surrounding pixels are immediately input to the Laplacian calculation unit 2, and the Laplacian calculation K f (D, , , −D, , −, , , ) + (D
,,,, -ri,,.

1) + (I)ffi,, -D,,,,yo) + (p,,,,
,-D,,. 1. . )) is carried out. Here, K is an arbitrary integer.

This value of the Laplacian calculation unit 2 is set as a reference threshold (for example, 12
7) is input to the binarization unit 3 as a binarization threshold. In the binarization section 3, the density data p0. . is compared with the binarization threshold TH, and (1) density data D0. . ＞TH
When , binary data is 0ffi. -255 (black) (2) Binary data 0ffi when density data D1, ≦TH. Binary data 07. -0 (white). . get.

This binary data 0111.11 is converted into 0.0 by the subsequent binary data output section 4. ,,, when -255 "]' (black) O□
When the value is 1, -0, it is output as “O” (white).

Further, the output of the binarization section 3 is input to the negative input of the arithmetic unit 5. The positive input of the arithmetic unit 5 contains the density data D7 of the pixel of interest,
. is included, the output of the calculator 5 (binarization error E,
, fi) is E,, -Dffi,, -0,,.

becomes.

Subsequently, the error distribution calculation unit 6 calculates a value for weight distribution of the binarization error E, 7 to unprocessed peripheral pixels using a predetermined weighting coefficient shown in the error diffusion matrix 7. Subsequently, the calculator 8 adds the weighted distribution value to the original density data to obtain new density data. The specific operation has been explained with reference to FIG. Now, before and after binarization error allocation, the density data of surrounding pixels is modified as follows.
F is given.

D□. +l −=D□,. 41+(K□□41/
Σ =, +) ・E□.

D, +1. R-1-Dm-1,-1'' (K,,,,,
, /Σ, , )−E, , .

=D,,,,+ (K, , , , /Σ, , )−E, .

-D　m-H,n. l+ Dffi. .

Do, 1 (K□a+fi++　/Σ,,I)　・Eo,.

However, for Σ,,,-,,,,,,+,,,n-,+
Nitsu-,,+nii. 1. . 41 By sequentially repeating the above-described sequence in the main scanning direction and the sub-scanning direction to calculate binary data, an output image representing a pseudo halftone can be obtained.

The above explanation will be specifically explained. For example, assume that the Laplacian calculation result obtained by the Laplacian calculation section 2 is 74 as shown in FIG.

Set the coefficient K to 0. If it is 5, then 0.5×74=37. Assuming that the fixed value determined by the density data bits is 127, the threshold value TH to be found is TH-1, 27
-37-90. The threshold value has been lowered, and areas that were previously judged as white are now judged as black.
The image is binarized in a direction that emphasizes the outline of a line drawing or the like.

Here, assuming that the density Dm,... of the pixel of interest is 100, 100 will be compared with the threshold value TH=90, so the binarization result of the binarization unit 3 will be "1". . Therefore, in this case, the output Em,n of the arithmetic unit 5 is 100-
255--155.

Note that "1" of the binarized output Oi is regarded as 255. Regarding this -155 as an error, an error diffusion calculation using error diffusion matrix 7 is performed, and the pixel of interest DIR,1
The error will be allocated to pixels after 1. In addition, 2
Valued output O4,. is “0”, the output E of the arithmetic unit 5
. . is the concentration data D7. .

Become that.

In the above embodiment, the present invention is applied to image data in a 3×3 matrix, but the present invention is not limited to this, and can be applied to image data in any other matrix. Can be applied.

Furthermore, in the present invention, an example is given in which a Laplacian operation is performed on density data that has been subjected to an error diffusion operation.
It is also possible to perform Laplacian calculation on original density data without error diffusion.

Furthermore, in the present invention, an example is given in which the Laplacian operation is performed using adjacent pixels on the front, rear, left, and right sides of a pixel, but the present invention is not limited to this, and it is possible to perform the Laplacian operation using a wider range of surrounding pixels. You can also do this.

Further, the calculation for determining the binarization threshold is not limited to the Laplacian calculation, and various other threshold determination calculations may be used. For example, calculations such as calculating the average value of the density of the entire pixel and applying this average value [weighting may be distributed depending on the position around one pixel] may also be used.

[Effects of the Invention] As described above in detail, according to the present invention, by adding the Laplacian calculation method to the error diffusion method, it is possible to obtain a good pseudo-halftone image.

[Brief explanation of drawings]

Fig. 1 is a flowchart showing the principle of the method of the present invention, Fig. 2 is a diagram showing an example of a Laplacian filter, Fig. 3 is an explanatory diagram of Laplacian operation, and Fig. 4 is a diagram showing an example of a circuit configuration for implementing the present invention. , FIG. 5 is an explanatory diagram of the DISA method, and FIG. 6 is an explanatory diagram of the error diffusion method. In FIG. 4, 1 is a density data line buffer, 2 is a Laplacian calculation unit, 3 is a binarization unit, 4 is a binary data output unit, 5 is a calculation unit, 6 is an error distribution calculation unit, and 7 is an error diffusion matrix. , 8 is an arithmetic unit.

Claims (3)

[Claims] (1) Perform Laplacian calculation between the pixel of interest and its surrounding pixels on the density data stored in the density data line buffer (step 1), and perform Laplacian calculation with a fixed value determined by the number of bits of the density data. Perform a difference calculation with the result, and use the difference calculation result as a threshold for binarization to binarize the pixel of interest (
Step 2) Performs a difference calculation between the density data of the pixel of interest and the data obtained as a result of binarization (Step 3) The result of the difference calculation is used as an error and applied to surrounding pixels using an error diffusion matrix. An image processing method characterized in that error distribution is performed by performing error diffusion calculation (step 4). (2) The image processing method according to claim 1, wherein the Laplacian operation is performed on original density data rather than on density data after error diffusion calculation. (3) The image processing method according to claim 1, wherein various threshold determination calculations are performed instead of the Laplacian calculation.