JPH048063A - Picture processing system - Google Patents

Abstract

PURPOSE:To obtain excellent picture quality from both a character and line part and a photograph and intermediate tone part by adopting the picture processing system through the use of the error diffusion method for binarization so as to obtain a pseudo intermediate tone picture thereby improving reproducibility of a character line drawing and correcting a delay in dot appearance at a picture where density is low. CONSTITUTION:A Laplacian arithmetic section 2 extracts a density data of a noted picture element and a density data of plural surrounding picture elements depending on a coefficient matrix adopted by the Laplacian arithmetic operation to apply the Laplacian arithmetic operation and the addition of the result of the Laplacian arithmetic operation and the data of the noted picture element. An error between an output binarized by a binarizing section 3 and the result of the Laplacian arithmetic operation by the Laplacian arithmetic operation section 2 is calculated and the result is fed to an error distribution arithmetic operation data correction section 5. Since a substantial gradation border is not in existence by utilizing the Laplacian arithmetic operation by the error spread method, excellent picture quality with emphasized contour is obtained from the character line drawing part and excellent picture quality with no noise produced therefrom is obtained from the photograph part.

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] This invention relates to an image processing method for obtaining a pseudo-halftone image by binarizing each pixel of an image composed of density data (each pixel has a plurality of vias).

The aim is to provide an image processing method that can obtain good image quality for both text/line drawings and photographs/halftones.The data of the pixel of interest and surrounding pixels is extracted from the density data memory and Laplacian calculation is performed using a predetermined coefficient matrix. The calculation result is added to the data of the pixel of interest in the Laplacian calculation unit, and the output of the Laplacian calculation unit is compared with a predetermined threshold value in the II conversion unit and binarized, and the value of the binarized result is The error calculation section calculates a binarization error from the result of the Laplacian calculation, and the error distribution calculation section distributes the binarization error value to the periphery of the pixel of interest.

[Industrial Application Field] The present invention relates to an image processing method for obtaining a pseudo-halftone image by binarizing each pixel of an image, each pixel of which is composed of multiple bits of density data, particularly for digital image processing. Device 2
For example, a digital copier.

The present invention relates to a halftone image processing method for obtaining pseudo-intermediate images in image scanners, facsimile machines, etc.

In recent years, in each of the above devices, each pixel has multiple values (2 or more)
A method (referred to as area gradation) is used in which each pixel of an image is binarized to obtain a halftone image. However, with conventional methods, the processed images have various drawbacks.
A solution is desired.

[Prior Art] In conventional devices that utilize various types of pseudo-intermediate image processing, a "systematic dither method" has been widely and commonly used as a halftone image processing method. This method of systematic dithering is
Since the hardware configuration is simple, it has the advantage of being able to express pseudo-halftones at low cost, but it has the following drawbacks.

(1) If the original is a halftone image from printing, etc., a periodic striped pattern II (referred to as moiré) that is not present in the original will appear on the processed image.
occurs.

(2) When a document contains characters, line drawings, etc., the processing results for those portions are cut off (-because the black pixels of the book's lines turn white at regular intervals), resulting in extremely poor reproducibility.

(3) In order to obtain multiple gradations, the dither matrix size is made too large relative to the resolution of the reading system.

If the resolution deteriorates and the matrix size is reduced, ′f
! #The number of tones is reduced, making it impossible to achieve both multiple gradations and high resolution.

On the other hand, as a halftone processing method that can achieve both multiple gradations and high resolution, there is a method called "error diffusion method." this is,
R, W, FIoyd and L, Steinber.
g”An Adaptive Algorithm
for 5patial Gray 5cal
e” 1975 SID International
l Symposiu+m Digest of Te
Chnical Papers, 4.3. pp36
-37 (Apr, 1975).

This error diffusion method is a method that is 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, and is similar to the dither method. The number of expressed gradations is not uniquely determined by the matrix size, and the gradation characteristics such as no moiré, etc.
Superior resolution.

Binarization processing using the error diffusion method will be briefly explained. Assuming that the pixels of the n-th line in the main scanning direction and the m-th line in the sub-scanning direction are to be binarized in the image space, the original image data representing the density of each pixel is calculated from the minimum density to the maximum density. It is assumed that the value is quantized into a value indicating the concentration and input.

First, the correction data (the sum of the original image data and the total error distribution value) corrected by the binarization error generated in the binarized surrounding pixels is compared with a predetermined threshold (intermediate density), and the binary (set to a value uniquely equivalent to the maximum density (black) or the minimum density (white)) to obtain binary data. Here, the difference generated during binarization is distributed as a binarization error to unprocessed surrounding pixels with a predetermined weight, and added to the original image data.

By repeating the above processing over the entire image space, binary data binarized by the error diffusion method is obtained, and by displaying and printing it, a pseudo intermediate image is obtained.

[tJHE to be Solved by the Invention However, the above error diffusion method also has the following problems.

■Reproducibility of characters and line drawings is poor compared to simple binarization processing.

■There is a spatial delay in the appearance of dots in areas where the concentration is low. In other words, a pixel with low density is binarized as "white", but the value of the difference between the original data of the pixel of interest and the binary data output as "white" is distributed to surrounding pixels, and the pixel with low density is Errors are similarly distributed for pixels of density, and a certain distance occurs before the accumulated errors reach a value that is converted into IF as "black".

SUMMARY OF THE INVENTION An object of the present invention is to provide an image processing method that can obtain good image quality in both text/line drawing areas and photograph/halftone areas.

In the present invention, correction data obtained by correcting original image data using processed pixels during processing using the error diffusion method is further corrected using a Laplacian decompression value based on the pixel of interest and its surroundings.

[Means to solve the problem] FIG. 1 is a diagram showing the principle configuration of the present invention.

In FIG. 1, 1 represents a density data memory, 2 a Laplacian calculation section, 3 a binarization section, 4 an error calculation section, 5 an error distribution calculation/data correction section, and 6 a binary data output section.

In a discrete space, the Laplacian operation is calculated as the sum of differences in density between the pixel of interest and surrounding pixels.

Various coefficient matrices (Laplacian filters) have been proposed. Correction using this Laplacian operation has the effect of emphasizing the outline of the image, and is generally used in the simple binarization method, and in the conventional systematic dither method, nine gradations change. It is not used because it has negative effects such as noise at the boundary.

[Operation] In FIG. 1, density data memory 1 stores density data for each dot of an original image (each dot is represented by a plurality of bits of density). The Laplacian calculation unit 2 takes out the density data of the pixel of interest and the density data of a plurality of surrounding pixels determined by the coefficient matrix employed in the Laplacian calculation, performs the Laplacian calculation, and adds the result of the calculation to the data of the pixel of interest. The calculation result of the Laplacian calculation unit 2 is supplied to the binarization unit 3, where it is compared with a set threshold value, and if it is equal to or greater than the threshold value, an output of “English” (black) is generated, and if it does not reach the threshold value, an output of “ 0” (white) is generated, and the binary data output unit 6 outputs it as binary data of the pixel of interest.
supplied to

This binary data is supplied to the error calculation section 4.

The error with the calculation result output from the Laplacian calculation unit 2 is calculated. In this case, when the output of the binarization unit 3 is “l”, the agricultural value of the concentration data is input to the error calculation unit 4, and the output is “0”.
”, the minimum value of the density data (all O) is supplied (because the binarization section determines the intermediate value as either “1” (black) or “0” (white)). be).

When the error calculation unit 4 calculates the error between the output binarized by the binarization unit 3 and the calculation result of the Laplacian calculation unit 2, the error value is supplied to the error distribution calculation/data correction unit 5. . The error distribution/data correction unit 5 performs a calculation to distribute the input error value to predetermined pixels (pixels to be processed from now on) among the pixels surrounding the pixel of interest at a preset ratio. , reads out the value of the corresponding pixel from the density data memory 1, adds the calculated distribution value to the read value, and writes the added value at the original position. Such processing is sequentially executed for each pixel.

Since the error diffusion method of the present invention uses Laplacian operation, there is no substantial gradation boundary, so good image quality with emphasized outlines can be obtained in single characters and line drawings, and noise is reduced in photographic areas. Good image quality is obtained.

[Embodiment] Fig. 2 is a block diagram of an embodiment of the present invention; Fig. 3 is a detailed block diagram of the embodiment; Fig. 4 is a configuration of a Laplacian calculation unit; and Fig. 5 is a configuration of an error distribution calculation unit. , FIG. 6 is an example of a Laplacian matrix.

In FIG. 2, reference numeral 11 corresponds to the density data line buffer (corresponding to 1 in FIG. 1), 12 to 13 correspond to FIG. is the binary data output section (first
17 represents an adder (15 and 17 together correspond to 5 in FIG. 1), and 18 represents an error diffusion matrix.

As shown in FIG. 6, various proposals have been made as coefficient matrices for Laplacian operations.

In the second embodiment, the upper, lower, and left sides of the pixel of interest are shown in (a).

Assume that the coefficients shown in the figure are used for the right pixel.

To explain the operation, 2. In an image reading device (not shown), an image is read in parallel with a line type image sensor etc. in which N (N is an integer) dots are lined up in the main scanning direction, and the original image data (minimum A 28-bit integer with a density of 0 and a maximum density of 255 is input. Then, this original image data is stored in the corresponding position of the density data line buffer 11. Note that this density data line buffer 11 is composed of a RAM (random access memory) or the like.

The pixel of interest is stored in the density data line buffer 11.

The pixel data of the n-th dot of the m-th line, that is, D
A case will be described in which @+R and the density data displayed in FIG. 2 are processed. Note that m and n are positive integers, and 1
≦n≦N.

The density data of the pixel of interest and its surrounding images are immediately input to the Laplacian calculation unit 12, where a Laplacian calculation is performed. This arithmetic expression is linear based on the coefficient matrix described above.

K・((D,,,-D,-,,,)+(D,,.

Da, e-+)+(Da,, -D,,,,,I)+(
D,,,D-.,,,)) Here, is any integer.

The value obtained by this calculation is added to the density data D0, 7, and the value is defined as correction data D@+T+°, and the binarization unit 1
Enter 3. In the binarization unit 13, the correction data D Il
l+11' to a predetermined threshold (for example, "127:255
(intermediate value of

■When correction data D,,,'>binarization threshold, iij element m
The binary data of n (01, represented by an o) is −255 (black).

■Correction data D @*n゛≦When the binarization threshold is small, it becomes 200 million data of 1 pixel mn (0,,,,,) -0 (white).

In this way, binary data OII+l+ of 9 pixels m and n is obtained. This binary data O1,, is the binary data output section 16
By, O,,, when -255, "l" (black), O
, s-0, it is output as "0" (white) and is also supplied to the adder 14. In the adder 14, the binarization errors E, 1. is required.

This binarization error E,, is supplied to the error distribution calculation unit 15, and the binarization error EI6+,, is given a predetermined weight by a coefficient (K1.0.A1~K@) according to the error diffusion matrix 18. * I + II + l 1, etc.) to allocate weight to unprocessed surrounding pixels. In this embodiment, as shown in FIG.
allocated to one pixel.

For distribution, extract the pixel data of the corresponding surrounding pixels,
This is performed by adding the values allocated to the pixel using the adder 17 and writing the addition result (data after allocation) at the position of the original pixel data.

The density data of surrounding pixels is corrected as follows before and after the binarization error is allocated.

By sequentially repeating the above operations in the main scanning direction and the sub-scanning direction to brighten the binary data, an image representing a pseudo halftone can be obtained.

In this example, the Laplacian calculation was performed using the image of interest and adjacent pixels on the front, back, left, and right sides, but the operation is not limited to this, and a wide range of surrounding pixels including (b) and (C) in Fig. 6 can be used. It can be implemented with other coefficient matrices. Similarly, the binarization error distribution range in the error distribution calculation unit 15 is not limited to the example shown in FIG. 2.

Next, a configuration embodying the present invention will be explained with reference to FIGS. 3 to 5.

FIG. 3 is a specific configuration diagram of the embodiment, FIG. 4 is a configuration of a Laplacian calculation section, and FIG. 5 is a configuration of an error distribution calculation section.

In FIG. 3, 31 and 32 are density data line buffers, 33 to 40 are latch circuits in which parallel input data is stored, 41 to 44 are adders, 45 is a Laplacian calculation unit, and 46 is an error distribution calculation unit. , 47 is a binarization circuit, and 48 is an adder for error calculation. Also, Figure 4,
In FIG. 5, 50 to 56 and 58 represent adders, 57 represents a conversion table, and 59 to 62 represent conversion tables.

In the configuration shown in FIGS. 3 to 5, when input density data is input in 8-bit length (representing density from 0 to 255),
It is assumed that each data line and each component has corresponding lines and elements for 8 bits or more.

In addition, the image space to be processed has N (for example, 172B) dots in the main scanning direction and M (for example, 23B) dots in the sub-scanning direction.
10) When composed of line pixels, the data line buffers 31 and 32 are for storing density data for N-3 pixels, and are composed of FIFO (first-in, first-out) type RAM. has been done.

The operations shown in FIGS. 3 to 5 will be explained below.

First, the input density data D i + n is sent to the latch circuit groups 33 to 40 . in synchronization with a transfer black signal (not shown). Adder 41
44 and the density data line buffers 31 and 32, and at the beginning of one image space,
All bits of the Q output of each latch circuit are reset to "0", and all bits of the Q output are reset to "1".

The Laplacian calculation section 45 performs calculations using a coefficient matrix similar to the embodiment shown in FIG.

In other words, the density data D of the pixel of interest (output of the latch circuit 37) and the one's complement data D of the density data of the four surrounding pixels, +111 D @+l'l-1+T5
m+*・ld,. InI3 (latch circuit 40.3
B, Q output of 36.34) as input data, adders 50 to 56 as shown in FIG. 4) is input to a conversion table 57 made up of a ROM (read only memory) and the like. In the conversion table 57, the result of multiplying by a predetermined coefficient is written in advance, and the Laplacian calculation result for the sum of density differences with two surrounding pixels is output.

Adder 58 adds this result and the density data D, A of the pixel of interest.
The density data D1. is added and corrected by D1. I get it.

Next, the corrected density data D, .

■D,, 11'> 0 output when THO ((L, s)' All pits O (8 bits) 0 output (
0,,,): 1 (1 bit)■Dll+ll゛≦
O output at THO (Os, l: All bits 1 (8 bits) O output (
0, , ): 0 (1 bit) In other words, in the case of ■, the binarized output (0, , ) of the target pixels m and n is "1", and the density data is the maximum value (all in focus 1). Then, the 1's complement of the maximum value becomes 0 for all bins as the O output. In the case of ■, the binarized output (0,,,) of the target pixels m and n is “0”
in.

The density data becomes the maximum value (all focus is 0).

The 1's complement of the maximum value becomes the O output and all the points are 1.

The binarization circuit 47 is composed of a comparator and a NOT circuit, and the O output (0, 15) and the corrected density data D@+II' are added in an adder 48, and 2 Sulfurization error data E11. is required.

Next, in the error distribution calculation section 46, the binarized error data E
, 1. Based on this, as shown in FIG. 5, the weighting of the error diffusion matrix (see 18 in FIG. 2) is calculated from the conversion tables 59 to 62, which store the error distribution determined in advance from the coefficients, for each unprocessed data. Error allocation value to pixels E11. Ya l l Ea*
+9m-11EII+I+II I El&$1+1.
1 is determined and set as input data for one of the adders 41-44. In addition, the conversion tables 59 to 62 in FIG.
is composed of ROM and the like.

The adders 41 to 44 output the Q outputs of the launch circuits 33 to 36, □1. D□11. l-11D□IIMID□1,
+al and the above error distribution value E 11+ll+I +
Em. 11, -+ Em41+Il + Em, 9.
. 1 and are added and output.

With the above operations, processing for one pixel is completed, and similarly, by a transfer lock signal (not shown), the processing is repeated for the entire image space by shifting by one pixel, and the binary data 0.1. Output sequentially.

In the configurations shown in FIGS. 3 to 5, the subtraction is realized by addition with a 1's complement (logical inversion output), but a +1 addition means may be added to obtain a 2's complement.

[Effect of the invention] According to the present invention, the error diffusion method is used to convert the signal into 2W.

The image processing method that obtains pseudo-halftone images improves the reproducibility of single characters and line drawings, and also corrects the delay in the appearance of dots in areas with low density, so it is possible to improve the reproduction of single characters and line drawings.
Good image quality can be obtained in both halftone areas.

[Brief explanation of the drawing]

Figure 1 is a diagram of the principle of the present invention, Figure 2 is a diagram of an embodiment, Figure 3 is a detailed diagram of the embodiment, Figure 4 is the configuration of the Laplacian calculation section, and Figure 5 is error distribution calculation. Composition of the section, 6th
The figure is an example of a Laplacian matrix. In Figure 1, 1: Density data memory 2 Nilabrasian calculation unit 3: Binarization unit 4: Error calculation unit 5: Error distribution calculation/data correction unit 6: Binary data output unit

Claims (1)

[Claims] In an image processing method for obtaining a pseudo halftone image by binarizing each pixel of an image, each pixel of which is composed of multiple bits of density data, A Laplacian calculation unit (2) performs a Laplacian calculation using a predetermined coefficient matrix and adds the data of the pixel of interest to the data of the surrounding pixels, and the output of the Laplacian calculation unit is converted to a binarization unit (3). Binarization is performed by comparing with a predetermined threshold value, and an error calculation unit (4) uses the value of the binarization result and the result of the Laplacian calculation.
An image processing method characterized in that a binarized error value is obtained by: and the binarized error value is distributed to the periphery of a pixel of interest by an error distribution calculation/data correction unit (5).