JP3256267B2 - Image processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus for performing halftone processing while preserving the resolution of characters.

[0002]

2. Description of the Related Art When an image is separated into image areas and adaptively binarized, the image area is generally separated into a character area on a white background and other areas. For example, MTF correction is performed on a character area. After that, binarization is performed with a fixed threshold, and other regions are subjected to dithering or binarization using an error diffusion method, thereby achieving both character resolution and pattern gradation (for example, electronic IEICE Technical Report IE9
See “Separation method of character / halftone / photo mixed image” described in 0-32.

However, it is technically difficult to accurately separate a character and a halftone dot existing on a halftone dot using a local processing technique that can be implemented by hardware. Even when characters are subjected to pattern processing and dither processing, as well as when binarization processing is performed by the error diffusion method, reproduced images of characters with shading on the original are unclear. There was a problem that it was difficult to read.

[0004] Therefore, a method of reproducing characters with a mesh on a manuscript with good resolution has been proposed (refer to the 10th Annual Meeting of the Institute of Image Electronics Engineers of Japan 1990). That is, a character edge is detected only with a simple contrast (that is, a portion having contrast is regarded as a character edge), and when it is determined as a character edge, error propagation in the error diffusion method is controlled with priority on character image quality. That is the method.

[0005]

However, in the above-described method, although the resolution of the character edge on a white background and a halftone dot is preserved, there are places where a character edge is frequently erroneously detected in a general halftone dot document. Due to the presence, the image quality of the picture is significantly deteriorated.

As another method, in order to improve the accuracy of character edge detection, in addition to the above-described contrast information, density information of a target pixel and information as to whether or not the target pixel is in the middle of a density gradient are obtained. A method has been proposed in which a character edge is forcibly output as black or white according to these results (see JP-A-3-18665). However, even if such a method is used, a part which is erroneously detected as a character edge in a general halftone dot document,
Although reduced compared to the previous method, there is still a problem that the image quality of the picture is degraded because it is still present.

As described above, conventionally, as long as it is determined whether or not image area separation is a character area by local processing, erroneous separation occurs, and as a result, deterioration of image quality due to erroneous separation is inevitable.

An object of the present invention is to provide an image processing apparatus which preserves the resolution of characters on a white background and characters on a halftone dot, and can reproduce a high quality picture.

[0009]

In order to achieve the above object, according to the first aspect of the present invention, a target pixel value of an input signal is obtained.
The threshold of each pixel whose pixel value has already been determined in the vicinity of the pixel of interest
Addition of weighted sum of error weights generated by processing
Calculating means, and thresholding the added value to generate a binary signal.
Output binarizing means, and each pixel of interest of the input signal
First calculating means for calculating the probability of being a character area;
Calculate the probability that each pixel of interest in the force signal is in the neighborhood of the character
Second calculating means, and each pixel of interest of the input signal is
Detecting means for detecting whether the area is an area or not;
The probability of being a character area calculated by the output means is equal to or less than a predetermined threshold.
When it is above, select the maximum density signal as the output signal,
It is certain that the area is the character vicinity area calculated by the second calculation means.
The rate is equal to or higher than a predetermined threshold, and a white background area is detected by the detection means.
When detected, select the minimum density signal and
Otherwise, select the binary signal and output signal
Means for selecting the number and the value calculated by the first calculating means.
The probability of being a character area, and calculated by the second calculating means.
The weighted sum is corrected according to the probability that the area is a character vicinity area.
And a correcting means .

According to the second aspect of the present invention, the probability of the character area is calculated based on the density level of a predetermined neighboring pixel including the target pixel and the contrast between the target pixel and the neighboring pixel of the target pixel. It is characterized by:

According to the third aspect of the present invention, the probability that the pixel is the character vicinity area is calculated based on a probability that the pixel of interest is a character area and a probability that a plurality of pixels near the pixel of interest is a character area. It is characterized by.

In the invention described in claim 4, the probability that the target pixel is a character neighboring area includes a probability that the target pixel is a character region, a probability that a predetermined pixel near the target pixel is a character region, and a predetermined distance from the target pixel. The pixel is calculated based on the probability that the pixel is a character area.

[0013]

The judging unit calculates a character degree P representing the probability that the pixel of interest is a black character edge and a character proximity degree Q representing the probability that the pixel of interest is near the black character edge. In the above case, the black character signal is output to the selection circuit,
When a white background pixel is detected when the degree of character proximity Q is equal to or greater than a predetermined threshold, a white background signal near a black character is output to the selection circuit. The selection circuit outputs a signal for hitting a dot in accordance with the instruction of the black character signal, and outputs a signal without dot in response to the instruction of the white background signal near the black character. The error addition coefficient calculation unit corrects the weighted sum of the errors according to the calculated values of P and Q.
That is, as P and Q are larger, the error is reduced, and a process that emphasizes the resolution is performed.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be specifically described below with reference to the drawings. FIG. 1 is a block diagram of an embodiment of the present invention. In FIG. 1, reference numeral 1 denotes a scanner which reads a document, converts the document into a digital signal of multiple gradations, and outputs the digital signal.
Is a determination unit that calculates a character degree P and a character proximity degree Q and detects a white background, and an error that calculates a value k to be multiplied by the sum of errors based on the determined character degree P and the character proximity degree Q. The additional coefficient calculator 4 is a binarizing circuit for binarizing a signal obtained by adding an error to the input multi-level signal with a fixed threshold, and the binarizing circuit 5 is based on the decision signal from the decision unit 2. Output from the conversion circuit,
A selection circuit for selecting the MAX signal and the MIN signal, 6 is an error buffer, and 7 is a weight mask.

In this embodiment, black characters on a white background and black characters on a halftone dot are processed with high resolution. FIG.
3 is a diagram showing a black character edge region (that is, a region inside several pixels of a black character) and a black character edge vicinity region.

FIG. 3 is a diagram showing a specific configuration of the judging section 2. The character degree P calculating circuit 21 and the character proximity degree Q calculating circuit 2 are shown in FIG.
2, a white background detection circuit 23, and a determination circuit 24. The character degree P is a value representing the probability that the pixel of interest is a black character edge. Here, the black character edge may be on any of a white background and a halftone dot.

In the following, a method of calculating the character degree P of the pixel of interest will be described. In a 3 × 3 mask, matching is performed for the four patterns shown in FIG.
Ask for. When Th1>Th2> Th3, the input signal is binarized by Th1 and Th2, and four types of pattern matching are performed on the binarized signal. That is, in the determination circuit 211, if (all pixels xi> Th1) is set as the character degree P1 from the result of the two pattern matchings.
then character degree P1 = 2else if (all pixels xi> Th2) then character degree P1 = 1else
Then character degree P1 = 0 is calculated.

The density level of the target pixel is L, the pixel A is
When the density level of A is A, the contrast calculation circuit 2
12 calculates the contrast (LA). Th4>
When Th5>Th6>Th7> Th8, the determination circuit 211 calculates the character level P2 of the pixel of interest using the calculated contrast.

If (LA> Th4) then character degree P2 = 5 else if (LA> Th5) then character degree P2 = 4 else if (LA> Th6) then character degree P2 = 3 else if ( L−A> Th7) then character degree P2 = 2 else if (LA> Th8) then character degree P2 = 1 The determination circuit 211 adds the calculated character degree P1 and character degree P2 to obtain four patterns. Is output as the degree of character P of the pixel of interest. The range of the value of the character degree P is
From 0 to 7 (3 bits), the larger the number, the higher the probability that the target pixel is a black character edge.

The degree of character proximity Q is a value obtained by calculating the probability that the pixel of interest is near the black character edge. The first calculation method is as follows: if (character degree P of attention pixel P ≧ Tha) then character proximity degree Q = 0 else if (character degree P of pixel of interest P <Tha) th
en Character proximity Q = maximum character P in a 5 × 5 mask centered on the pixel of interest. That is, the character proximity Q is
Since the degree of surroundings of the character is indicated, the degree of character P is low.

The second calculation method is as follows: if (character degree P of attention pixel P ≧ Tha) then character proximity degree Q = 0 else if (character degree P of pixel of interest P <Tha) th
en Character proximity Q1 = Maximum character P among 3 × 3 masks centered on the pixel of interest, character proximity Q2 = 5 × 5 mask shown in FIG. 5 centered on the pixel of interest (this mask is From the maximum character degree P at the weight 1 portion of the mask (the mask having a larger weight closer to the pixel of interest), the character proximity degree Q becomes the character proximity degree Q = MAX (2 × Q1, Q2). Character proximity Q
The calculation circuit 22 performs the above-described filter operation using the character degree P, and outputs a character proximity degree Q.

In the following description, the character proximity Q obtained by the second method will be used. Here, the black character edge may be on any of a white background and a halftone dot. The range of the value of the character proximity degree Q is 0 to 14 (4 bits).
The pixel of interest has a high probability of being near the black character edge.

The white background detection circuit 23 detects a white background near the target pixel. That is, in the binarization circuit, pixels having a predetermined threshold value or less are output as white pixels with respect to the input multi-level signal. At this time, the MTF correction filter shown in FIG. 6 may be applied to the input signal. Next, in the 5 × 5 mask centered on the pixel of interest, if there are basically 5 × 2 white pixels as shown in FIG. 7, the pattern matching circuit detects the pixel of interest as a white pixel. , 5
If there is at least one white background pixel in the × 5 mask, the center pixel is output as a white background pixel.

The determination circuit 24 outputs a black character signal / a white signal in the vicinity of a black character based on the character level P, the character proximity level Q, and the white background detection. That is, in the pixel of interest, when the character degree P> T0, the black character signal is turned on (upper bit). When the character proximity degree Q> T1 and the pixel is white, the white background signal near the black character is turned on (lower bit). , T
The threshold values of 0 and T1 are determined by simulation.

The error addition coefficient calculation unit 3 calculates a value k to be multiplied by the sum of the errors based on the character degree P and character proximity degree Q signals from the determination unit 2. Tables as shown in FIGS. 8 and 9 are prepared, and either k [P] or k [Q], whichever is greater, k
= MAX (k [P], k [Q]) is selected. The larger the character degree P and the character proximity degree Q are, the larger the pixel of interest is.
There is a high possibility that the character is at or near the character edge. In this case, the error is reduced to obtain a reproduced image that emphasizes the resolution. That is, the value k by which the sum of the errors is multiplied is made close to zero.

The selection circuit 5 outputs an output from the binarization circuit 4 based on the black character signal / white background signal in the vicinity of the black character from the determination unit 2, a signal (MAX) for dot formation, and no dot formation. (MIN) is selected. That is,
When the black character signal is on, the selection circuit 5 selects MAX. When the white background signal near the black character is on, the selection circuit 5 selects MIN. Otherwise, the output from the binarization circuit 4 is selected. select.

The binarizing circuit 4 binarizes a signal obtained by adding an error to an input multi-level signal with a fixed threshold. Also,
In order to reduce the texture peculiar to the error diffusion method, for example, a halftone type dither matrix may be used, and the threshold may be binarized with periodicity. FIG. 10 shows an example of a mask used as the error weight mask 7. The weight of the mask is not limited to this, and its weight can be changed according to the image quality.

The operation of the embodiment constructed as described above will be described. In the error diffusion method, as shown in FIG. 1, the density x (i, j) of a pixel of interest in an input image Weight w of the error e (i, j) generated by the threshold processing of each pixel y (i, j) of the binary image whose pixel value has already been determined.
Correction value f to which the weighted sum E (i, j) of (i, j) is added
This is a method of obtaining a binary image by taking (i, j) and binarizing the correction value with a threshold th. That is, y (i, j) = th {f (i, j)} th (f) = 1 (when f> th), 0 (otherwise) f (i, j) = x (i, j) ) + E (i, j) E (i, j) = Σw (n, m) e (n, m) e (n, m) = f (n, m) −y (n, m)

On the other hand, in the binarization process by the error diffusion method of the present embodiment, E (i, j) is corrected.
That is, f (i, j) = x (i, j) + kE (i, j), and the weighted sum E (i, j) is calculated by the error addition coefficient calculation unit 3
Is corrected by the k value calculated in step (1).

Therefore, the higher the degree of the character, the lower the k value becomes.
And the binarization error becomes zero. At this time, the determination unit 2 outputs a black character signal or a white background signal near the black character, and the selection circuit 5 outputs a maximum density signal (MAX).
Alternatively, a minimum density signal (MIN) is output. on the other hand,
As the degree of the character is lower, the k value approaches 1, and a process that corrects the binarization error and emphasizes the gradation is performed.

In the above-described method of calculating the character degree P and the character proximity degree Q, there is a pixel region having a high probability of being a black character edge in a halftone dot image. Thus, the black character edge probability in a halftone dot image is corrected by using the method of detecting a halftone dot region by detecting a peak pixel as described in the above-mentioned paper “Separation method of character / halftone / photo mixed image”. It is also possible. FIG. 11 shows a configuration of a determination unit to which a dot area detection unit is further added.

That is, if the peak pixel density is extremely high in the vicinity of the pixel of interest, the degree of character P and the degree of character proximity Q may be reduced. Specifically, as described in the article,
The peak pixel is detected with a 3 × 3 mask, and the peak pixel is counted with a predetermined size centered on the target pixel, for example, an 8 × 8 mask. Judge, and output the signal “0”; otherwise, output the signal “1”. As a result, the character degree P and the character proximity degree Q of the halftone dot area are corrected to small values.

As described above, according to this embodiment, different processing is performed between the vicinity of the character on the white background and the vicinity of the character on the halftone dot, so that the resolution of the character on the white background and the character on the halftone dot are preserved. In addition, since the processing is continuously switched based on the probability of the character degree P and the character proximity degree Q indicating that the target pixel is likely to be located, discontinuity at the time of processing switching is inconspicuous. In particular, erroneous separation points in a halftone dot document become less noticeable.

[0034]

As described above, according to the present invention , since ambiguous separation results such as the degree of character and the degree of character proximity are introduced, character priority processing (processing emphasizing resolution) is replaced with picture priority processing. (Processing emphasizing gradation) can be continuously changed. Therefore, even when the character priority processing is selected in the halftone original, the deterioration of the image quality is not noticeable, and even when the character priority processing is selected at the edge portion of the photo original, the deterioration of the image quality is not noticeable. In addition, since a process that takes into account the appropriate error is applied around the characters other than the white background without putting dots on the white background around the characters, sufficient contrast is given to the characters and the background, and The result is an image with no images. In addition, the amount of hardware in this embodiment can be made smaller than a device for separating an image area plus a device configuration using an error diffusion method.

According to the present invention , the probability that the target pixel exists at the character edge can be calculated by a simple circuit.

According to the present invention , the probability that the target pixel exists near the character edge can be calculated by a simple circuit.

[Brief description of the drawings]

FIG. 1 is a block diagram of a first embodiment.

FIG. 2 is a diagram showing a black character edge region and a black character edge vicinity region;

FIG. 3 is a diagram illustrating a specific configuration of a determination unit.

FIG. 4 is a diagram showing four types of patterns for calculating the degree of character.

FIG. 5 is a diagram illustrating an example of a weighting mask for calculating the degree of character proximity.

FIG. 6 is a diagram showing a filter for MTF correction.

FIG. 7 is a diagram showing a pattern for detecting white pixels.

FIG. 8 is a diagram showing a table storing k values corresponding to character degrees.

FIG. 9 is a diagram showing a table in which k values corresponding to the degree of character proximity are stored.

FIG. 10 is a diagram showing an error weight mask.

FIG. 11 is a diagram showing a configuration of another embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Scanner 2 Judgment part 3 Error addition coefficient calculation part 4 Binarization circuit 5 Selection circuit 6 Error buffer 7 Weight mask

Continuation of the front page (56) References JP-A-63-288567 (JP, A) JP-A-62-139472 (JP, A) JP-A-3-85968 (JP, A) JP-A-5-122510 (JP) , A) JP-A-63-214073 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) H04N 1/40-1/409 H04N 1/46 H04N 1/60

Claims (4)

(57) [Claims] 1. An adding means for adding a weighted sum of weights of errors generated by threshold processing of each pixel whose pixel value has already been determined in the vicinity of a target pixel to a target pixel value of an input signal, and A binarizing unit that outputs a binarized signal by performing a threshold process; a first calculating unit that calculates a probability that each pixel of interest of the input signal is a character area; A second calculating means for calculating the probability of being an area; a detecting means for detecting whether each target pixel of the input signal is a white background area; and a character area calculated by the first calculating means. When the probability is equal to or greater than a predetermined threshold, the maximum density signal is selected and output as an output signal, and the probability of the character vicinity area calculated by the second calculation means is equal to or greater than a predetermined threshold, and the detection means selects a white background. When an area is detected,
Selecting means for selecting the minimum density signal as an output signal, and otherwise selecting the binarized signal as an output signal; and a probability of being a character area calculated by the first calculating means, An image processing apparatus comprising: a correction unit that corrects the weighted sum according to the probability that the region is a character vicinity area calculated by the second calculation unit. 2. The method according to claim 1, wherein the probability of the character area is calculated based on a density level of a predetermined neighboring pixel including the target pixel and a contrast between the target pixel and the neighboring pixel of the target pixel. Item 2. The image processing apparatus according to Item 1. 3. The probability of being a character vicinity area is calculated based on a probability that a pixel of interest is a character area and a probability that a plurality of pixels near the pixel of interest is a character area. 2. The image processing device according to 1. 4. The probability that the pixel of interest is a character region, the probability that a predetermined pixel in the vicinity of the pixel of interest is a character region, and the probability that a pixel that is a predetermined distance from the pixel of interest is a character region. 2. The image processing apparatus according to claim 1, wherein the calculation is performed based on a certain probability.