JP2962742B2 - Image processing device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus for accurately separating a character area in an image, particularly a character area existing on a halftone dot.

[Conventional technology]

Generally, in an image called a document image, characters (including line drawings), photographs, halftone dots, and the like are usually mixed.
When such an image is reproduced by, for example, a digital copying machine or a facsimile, (a) in order to improve the image quality of the reproduced image, it is reproduced clearly so that characters and lines are not cut off.

(B) Emphasis is placed on the gradation of photographs and halftone dots, and moire is removed from halftone dots.

It is desirable to take such measures. For example, in the case of reproducing an image in monochrome binary, for example, in order to realize the above (a), the edges of characters and lines are emphasized, and then binarization processing is performed.

 In order to realize the above (b), dither processing is performed after smoothing a photograph or a halftone dot.

And the like.

By the way, in order to perform the above-described processing, it is necessary to first separate a character region in an image from other regions other than characters (hereinafter, referred to as a non-character region) before the processing. Conventionally, as one of the separation methods, for example, a method proposed by Sato (“Moiré removal processing by dither method”, Proceedings of the 2nd Non-impact Printing Technology Symposium, 3
-4, PP. 69-72, 1985). In Sato's method, input data is first smoothed in order to eliminate a halftone dot edge, the smoothed data is edge-emphasized, and then binarized to make a portion having a high threshold value a character area. .

[Problems to be solved by the invention]

By the way, characters and lines in an image do not always exist on a white background. For example, as shown in FIG. If the above-described processing is to be performed on the image shown in FIG. 3A, the character area surrounded by the dotted line (the character "i" and an area having a width of several dots around it) and the other non-character areas are separated. After separation, it is necessary to adaptively perform the above-described processing on each area as shown in FIG.

However, in the case of Sato's method described above, when the input data is smoothed, not only the halftone dot edges but also the character edges are erased at the same time. This tendency is particularly strong in the case of small characters or characters with thin lines. For this reason,
There is a problem that a character area cannot be accurately separated depending on an input image.

The present invention has been made in order to solve the above-described problem, and it is possible to accurately separate a character area in an image not only when a character exists on a white background but also when it exists on a halftone dot. It is an object of the present invention to provide an image processing apparatus that can perform the processing.

[Means for solving the problem]

In order to achieve the above object, the present invention provides a smoothing means for smoothing input pixel data with a predetermined weighting coefficient, and a predetermined predetermined amount on both sides of a pixel of interest of the pixel data smoothed by the smoothing means. When the density of the pixel at the position is lower than the density of the pixel of interest by a predetermined value or more, a ridge pixel detecting means for detecting the pixel of interest as a ridge pixel; and a ridge pixel for a predetermined two-dimensional area composed of adjacent pixels. A character region determining unit that determines that the two-dimensional region is a character region when the number of ridge pixels detected by the outgoing detection unit is a predetermined number or more.

[Action]

After converting the input image signal to digital multi-level data,
Smoothing is performed with a predetermined weight coefficient. As this smoothing filter, for example, And so on. The coefficient k is about 1 to 10, and an optimum value is selected according to the size of the character to be separated, the number of halftone dots, and the scanner or printer used.

When the above-described smoothing process is performed, the density level of a halftone dot composed of isolated small dots is reduced, and the density level difference between the halftone dot and the character becomes larger, so that the character region and the non-character region are easily separated. However, as described above, the edges of small characters and characters with thin lines are easily rounded due to smoothing, and if the smoothed digital multi-valued data is used as it is, character regions and non-character regions can be accurately separated. Is difficult.

Therefore, according to the present invention, for the smoothed digital multi-valued data, a pixel forming a character in an image is detected based on a density change in a predetermined local region.
It is desirable to detect a ridge pixel in which character information is relatively stored even by smoothing as a pixel forming a character.

Briefly describing the ridge pixel with reference to FIG.
(A) The figure shows the image of the character "i", and the thick solid line in the figure is the so-called ridge pixel with the highest density level, and the hatched area around it indicates the peripheral pixels connected to the ridge pixel. . Therefore, (a) shows the density level (black density) of each pixel along the line AA in the figure, as shown in (b), and the pixel at the density level maximum position indicated by hatching is the ridge pixel. give.

The ridge pixel is detected by comparing a local two-dimensional pattern of the smoothed digital multi-valued data with a predetermined ridge pixel pattern prepared in advance. That is, ridge pixel pattern with local region of a predetermined size, and the density level of the center pixel in the region and L _C, and said center pixel L _C is sandwiched in an arbitrary pixel pair concentration level in the symmetrical position when each L _a, and L _b, when Naritate below in the local area (2) is any one, and extracts the center pixel L _C as ridge pixels.

L _c -L _a ≧ TH and L _c -L _b ≧ TH TH: threshold - (2) shows an example of a concrete ridge pixel pattern in Figure 3. In this example, a ridge pixel pattern has an M × M pixel size (M =
5) the matrix obtained by the adoption of, in this matrix, for example L _c -L ₁ ≧ TH and L _c -L ₂₄ ≧ TH or L _c -L ₃ ≧ TH and L _c -L ₂₂ ≧ TH or L _c When at least one of the four pixel pairs of −L ₅ ≧ TH and L _c −L ₂₀ ≧ TH or L _c −L ₁₁ ≧ TH and L _c −L ₁₄ ≧ TH holds, the target pixel L _{c is set} to the ridge. It is a pixel.

Note that, when viewed locally, characters are often connected linearly in one direction at a certain density level. Therefore, if the target pixel _Lc is a ridge pixel when Expression (2) is satisfied for three of the four pixel pairs,
More precisely, only the ridge pixel of the character can be extracted. For example,
In FIG. 3, if the ridge of a character is present along the upper pixel _{L 3 -L 8 -L 17 -L 22} , 3 pixels of L ₁ and L _24, L ₁₁ and L _14, L ₅ and L ₂₀ Since equation (2) is satisfied in the pair, the target pixel
L _c may be the ridge pixel of the character.

The ridge pixel detected as described above may be determined as a character region as it is, but by performing the following determination processing, the character region can be separated with higher accuracy.

That is, for the ridge pixels obtained as described above, a predetermined local area, for example, a block of N × N pixel size is adopted, and the number of ridge pixels P is counted in units of this block. The size N of the block is determined according to the reading density of the image, the size of the character to be separated, and the like.

Then, based on the counted ridge pixel number P, it is determined whether or not the target pixel in the block is a character area. The determination of the character area employs one of the following determination conditions depending on whether the block is moved while moving the block in units of one pixel or while moving the block in units of one block.

(1 case moves pixel-by-pixel basis) count P of the ridge pixels for a given threshold value P ₀ or more, and the center pixel character region in the block.

(1 if moving block at a time) the case count P of the ridge pixels is equal to or greater than a predetermined threshold value P _0, and all pixels a character area in the block.

〔Example〕

FIG. 1 shows an image processing apparatus 1 according to the present invention.
Here is an example.

The input image signal unit 1 is a circuit that raster scans an original image and takes in the digital multivalued data. If the reproduced image is monochrome, a luminance signal is input. If the reproduced image is color, an RGB signal after color separation or a YMC signal after color correction is input. In the case of color, the processing described below may be performed in parallel for each color signal.

The smoothing processing unit 2 is a circuit that smoothes the digital multi-valued data sent from the input image signal unit 1 using a predetermined weighting factor represented by the equation (1).

The ridge pixel detecting unit 3 applies a ridge pixel pattern of M × M pixel size shown in FIG. 3 to the digital multi-valued data smoothed by the smoothing processing unit 2, and calculates the character This circuit detects a ridge pixel.

The character area determination unit 4 counts the ridge pixels detected by the ridge pixel detection unit 3 in N × N pixel size blocks, and determines from the count value P whether or not the target pixel in the block is a character area. It is a circuit for determining.

The operation of the above embodiment will now be described with reference to the flowchart of FIG. The character area determination unit 4 adopts the block of the N × N pixel size as the local area,
It is assumed that the character area is determined while moving in block units.

The image signal is converted into digital multi-value data in the input image signal unit 1 and stored. The digital multi-valued data is sent to the smoothing processing unit 2 and is given by the equation (1) and smoothed by a predetermined weight count (step [1]).

The ridge pixel detector 3 reads the smoothed digital multi-valued data for each M lines (step [2]), and applies an M × M pixel size matrix shown in FIG. 3 to the digital multi-valued data. and, the center pixel L _c of the matrix to detect whether the ridge pixels based on the (2) equation (step [3]).

When the detection of the ridge pixel is completed, the character area determination unit 4
Sets ridge pixel data by N lines (step [4]), counts the number of ridge pixels P in a block of N × N pixel size (step [5]), and obtains the obtained ridge in step [6]. comparing the number of pixels P with a predetermined threshold value P _0.

For P ≧ P _0, determining that all pixels in the block is a character area (step [7]), and outputs the character region signal. On the other hand, in the case of P <P _0, the pixel in the block is determined to be a non-character area (step [8]), and outputs the non-character region signal.

FIG. 5 shows an example of a copying machine constructed using the above-described character area separating apparatus as the image processing apparatus of the present invention.

In the figure, reference numeral 5 denotes a character area separating apparatus which is the image processing apparatus of the present invention shown in FIG. 6 is a sharpening processing circuit for edge enhancement for character area processing, 7 is a binarization processing circuit, 8 is a smoothing processing circuit for non-character area processing, and 9 is an emphasis on tonality for producing halftones. A swirling type dither processing circuit 10 is an image signal selection circuit for selecting either the binarization processing circuit 7 or the dither processing circuit 9 according to the character / non-character area signal output from the character area separation device 5. The above processing is realized by the sharpening processing circuit 6 and the binarization processing circuit 7, and the above processing is realized by the smoothing processing circuit 8 and the dither processing circuit 9, respectively.

The image signal is input in parallel to each of the character area separating device 5, the sharpening circuit 6, and the smoothing circuit 8. The character area separating device 5 determines whether or not all pixels of the input image signal are character areas as described above, and sends the determination result to the image signal selection circuit 10.

The image signal selection circuit 10 selects the binarization processing circuit 7 when the character area signal is output from the character area separation device 5, and selects the dither processing circuit 9 when the non-character area signal is output. Choose the side. As a result, in a character area in the input image, an image sharpened by the sharpening processing circuit 6 and the binarization processing circuit 7 is output. In a non-character area, the smoothing processing circuit 4 and the dither processing circuit 9 are output. The image subjected to the pseudo halftone processing is selectively switched and output.

Therefore, if the input image is copied and reproduced by using the image signal output from the image signal selection circuit 10, the character area is sharpened, and the non-character area is pseudo-halftone processed in a high-quality two-dimensional image. A value image is obtained.

Note that the smoothing processing circuit 4 in FIG. 7 and the smoothing processing unit 2 in FIG. 1 can be shared.

〔The invention's effect〕

According to the character separation method described in claim (1), it is possible to accurately separate a character region in an image not only when a character exists on a white background but also when a character exists on a halftone dot. it can.

In the case of the character separation method according to claims (2) and (3), the separation accuracy of the character area can be further improved.

[Brief description of the drawings]

FIG. 1 is a diagram showing an embodiment of the present invention, FIG. 2 is an explanatory diagram of a ridge pixel, FIG. 3 is a diagram showing an example of a ridge pixel pattern, FIG. 4 is a flowchart of the above embodiment, FIG. FIG. 6 is a diagram showing an example of a copying machine using the character area separating apparatus of the present invention. FIG. 6 is a diagram showing an example of reproducing a character image on a halftone dot.

──────────────────────────────────────────────────続 き Continued on front page (58) Field surveyed (Int.Cl. ⁶ , DB name) H04N 1/40-1/409

Claims (2)

(57) [Claims] 1. A smoothing means for smoothing input pixel data by a predetermined weighting coefficient, and a density of pixels at predetermined positions on both sides of a pixel of interest of the pixel data smoothed by the smoothing means. When the density is lower than the density of the target pixel by a predetermined value or more, the ridge pixel detection unit detects the target pixel as a ridge pixel, and the ridge pixel detection unit detects a predetermined two-dimensional region composed of adjacent pixels. An image processing apparatus comprising: a character area determining unit that determines that the two-dimensional area is a character area when the number of ridge pixels is equal to or more than a predetermined number. 2. The method according to claim 1, wherein the ridge pixel detection means detects the ridge pixel by determining that the density of pixels at predetermined positions on both sides in the vertical, horizontal, and tilt directions around the pixel of interest is lower than the density of the pixel of interest by a predetermined value or more. 2. The image processing apparatus according to claim 1, wherein the determination is made as to whether the target pixel is detected as a ridge pixel when there are a plurality of determinations that the target pixel is lower than a predetermined value.