JP2729278B2 - Image binarization expression method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image binarization expression method, and particularly to an improvement in an error diffusion method.

[Conventional technology]

Table 1 shows a summary of the conventional binarization expression method of images.

Although the purpose of the binarized representation of an image can be generally reduced in the amount of information, some of the features of the original image are necessarily lost due to the reduction in the amount of information. For example, when the original image is an image in which a graphic part such as a character or a drawing should be clearly displayed, binarization (narrow sense) is performed so that the graphic part and the background can be separated, so that the main part of the original image is Features are preserved. On the other hand, when the original image is an image to be expressed by a gray level, such as a three-dimensional natural image, a pseudo multi-tone expression should be used by dithering or the like. The shading characteristics are lost. However, in the printing field, the facsimile field, and the like, an image including both characters and photographs is often processed, and the characteristics of the entire image cannot be faithfully reproduced by a single binarized expression method.

Therefore, based on the differential value distribution of the image, a character region (a region in which binarization in a narrow sense is to be performed) and a graphic region (a region in which pseudo multi-tone expression is to be performed) are divided into regions. There has been proposed a method of performing binarization and dithering a graphic region. In such complex processing, the boundary between the two regions becomes discontinuous, and the result of the binarized expression becomes an extremely unnatural image.

On the other hand, as another binarization expression method, there is an error diffusion method, and a pseudo-multi-tone expression without limitation on the number of gradations is possible.
Here, the error diffusion method will be described.

The operating principle of the error diffusion method is as shown in FIG.
The pixel value (for example, luminance) at the coordinates (m, n) of the original image is represented by f
When the _mn, in which binarizes the f _mn taking into account the influence of the binarization error of the surrounding pixels. For example, let f _mn be the threshold R / 2
Binarized in, when to perform the conversion of the _{f mn ≧ R / 2 → R} f mn <R / 2 → 0, the pixel f ₀₀ of the first _{(0,0), e 00 = f} 00 [f 00 −R / 2] (1) Here, [] is a Gaussian function.

For a general pixel _fmn , a predetermined surrounding area (six pixels including the target pixel (x mark) in FIG. 1) is defined, and each error in this area is weighted (w _{1 in} FIG. 1).
~w ₆₎ to define. The weight matrix for the surrounding pixels is called an error filter. Binary error e for pixel f _mn
mn is calculated using the error filter shown in FIG. Is given. here, It is defined as Therefore, equation (2) can be rewritten as follows.

e _mn = _mn− g _mn (5) As described above, the binarization error e _mn includes the integrated binarization error in the pixels near the target pixel, and the luminance of the entire binarized representation image is The brightness of the entire original image is made as close as possible, and the original image and the binarized expression image are also approximated as much as possible with respect to the density distribution. The binarization error of each pixel is stored in an error buffer. However, the characteristics of the binarized representation image of the error diffusion method are determined by the error filter, and it is unlikely that both the character region and the graphic region are optimally represented.

Further, as an improvement of the error diffusion method, there has been proposed a method of removing a stripe pattern caused by the error filter and increasing the likelihood of a black point by using the error filter as a look-ahead type. However, as far as the processing result of this method is concerned, an unnatural image in which the outline is extremely enhanced occurs.

[Problems to be solved by the invention]

The present invention has been made in order to solve such a conventional problem, and can generate an image which faithfully reproduces the characteristics of the entire image regardless of a character area and a graphic area and gives a natural impression. It is intended to provide a valuation expression method.

[Means for solving the problem]

In the binarization expression method according to the present invention, a value obtained by multiplying a binarization error in a pixel near a pixel to be processed by a weight is added to the pixel value of the pixel to be processed, and a binary value of the image to be subjected to threshold processing is obtained by adding this addition result. In the generalized expression method, the weight is changed according to the pixel value of the processing target pixel.

[Action]

According to the binarization expression method according to the present invention, since the weight of the binarization error is changed according to the pixel value, each of the character area, the graphic area, and the background area is appropriately binarized and expressed. Are smoothly connected to generate an image giving a natural impression.

〔Example〕

Next, a first embodiment of a binarizing expression method of an image according to the present invention will be described with reference to the drawings.

In FIG. 2, the error filter of this embodiment includes a pixel corresponding to the pixel to be processed and a pixel corresponding to the pixel to be processed in the pixel immediately before the pixel to be processed and the scan line immediately before the scan line of the pixel to be processed. It consists of front and rear pixels. In FIG. 2, symbols e ₁ to e ₄ are assigned to the binarization errors of these pixels along the scanning direction. Then, the error filter whereas binarization error of _{e 1 ~e 4, w 1 ~w} 4
Weight.

These weights w _{1 to} w ₄ are not constant values but change according to the pixel value of the processing target pixel in the original image. Thus when the pixel value f _mn, weights w ₁ to w ₄ as a function of the pixel value _{f mn, w i (f mn} ) (i = 1~4) is expressed as (6).

Here, using the same expression method as the conventional error diffusion method,
This embodiment is expressed as follows.

Expression (7) is not limited to i = 1 to 4, but uses an expression applicable to an arbitrary error filter.

Here, the relationship between the weight w _i (f _mn ) and the binarized expression image will be considered.

If w _i (f _mn ) = 0, then _mn = f _mn and g _mn = R [f _mn −R / 2], which is equivalent to binarization in a narrow sense.

On the other hand, when w _i (f _mn ) >> 0, the surrounding binarization error strongly reflects on the pixel to be processed, and when there are many white pixels around as a result of the binarized expression, the pixel to be processed is reversed. It tends to be black pixels. Therefore, when viewed in a certain area unit, the image has an averaged density distribution, and the same effect as so-called smoothing occurs. This is suitable for smooth halftone expression.

From the above, for the character area, w _i (f _mn ) is reduced,
By increasing w _i (f _mn ) for the graphic area,
The feature can be faithfully reproduced for the entire image.
Furthermore, since the weight is determined for each pixel based on the pixel value and can be changed for each pixel, the respective regions are smoothly connected.

There are various possible characteristics of w _i (f _mn ).
Good results are respectively obtained by the characteristics shown in FIGS.

Figure 3, for w ₁ to w _4, are those which give a tendency of monotone increase with respect to f _mn, which increases the weight higher the pixel value is large.

For example, in an image including a halftone area ranging from intermediate luminance to high luminance and a character having low luminance (for example, black), if such weighting characteristics are given, the character area is sharply binarized, and the halftone area , An appropriate multi-tone expression is performed.

FIG. 4 shows a characteristic in which the luminance is monotonically increased below the luminance R / 2 and monotonically decreased above the luminance R / 2. In the low luminance region and the high luminance region, clear narrow binarization is performed. Is expressed in multiple gradations.

As a result, sharp binarization is performed on low-luminance characters and high-luminance backgrounds (for example, white).

Conventionally, in the background area, monochromatic expression has not been often performed, and the dither pattern has a strong tendency to be scattered. In this case, not only is the display of characters and graphics unclear, but also the data amount in the background area is significantly increased, and the communication efficiency of the facsimile is significantly reduced. On the other hand, by performing binarization in a narrow sense or a process similar to the binarization in the background region, it is possible to display the background region in a single color.

FIG. 5 shows that the characteristic curves on both sides have a convex upward tendency in addition to the tendency of the left-hand monotone increase and the right-hand monotone decrease of FIG. As a result, the region of multi-tone expression is expanded, and binarization in a narrow sense is performed only in the region of extremely low luminance and extremely high luminance.

FIG. 6 shows that the left characteristic curve has a tendency to change from a downwardly convex curve to an upwardly convex curve in addition to the tendency of the left monotone increase and the right monotone decrease of FIG. Gives a tendency to change from an upwardly convex curve to a downwardly convex curve. According to such a bell-shaped characteristic, the sharpness of the narrow-sense binarized region is further enhanced, and the smoothness of the multi-tone expression region is further enhanced.

Note in the error filter of FIG. _{2, w 1 = w 2 = w} 3 1.5w 1 ≦ w 4 ≦ 2.5w have obtained good results in terms of ₁ 0.10 ≦ w1 ≦ 0.26, typically an intermediate value Has been found to be a good choice.

FIG. 7 is a block diagram showing a second embodiment of the present invention. The setting of the weight of the error filter is performed based on the differential value of the original image. The differential value tends to be remarkably large at a graphic boundary, and in a halftone region such as a graphic region, the differential value tends to be large compared to the character portion of the character region and the background region. Therefore, it is of course possible to adjust the weight based on the differential value (first-order derivative, second-order derivative, Laplacian, Sobert operator, and other derivative operators).

〔The invention's effect〕

As described above, according to the binarization expression method according to the present invention, the weight of the binarization error is changed according to the pixel value, so that each of the character area, the graphic area, and the background area can be appropriately adjusted to the binary value.
This is an excellent effect that the image is expressed in a value and the boundaries between the regions are smoothly connected to generate an image giving a natural impression.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the operation principle of a conventional error diffusion method, FIG. 2 is a block diagram showing a first embodiment of a binarization expression method according to the present invention, and FIG. 4 to 6 are graphs showing the first characteristic of FIGS.
FIG. 7 is a block diagram showing a second embodiment of the present invention.

Claims (1)

(57) [Claims] 1. A value obtained by multiplying a binarization error in a pixel near a processing target pixel by a weight to a pixel value of the processing target pixel,
In a binarization expression method for an image in which this addition result is subjected to threshold processing, the weight is changed according to a pixel value of a pixel to be processed.