JPH06150059A - Image area separator - Google Patents

Abstract

PURPOSE:To improve separating accuracy by judging the concentrated part of lattice patterns and lines as a character area. CONSTITUTION:An edge area detection part 1 detects parts, where white picture elements are continued and black picture elements are continued, and outputs a place, where both of those parts are existent, as an edge area. A dot area detection part 2 detects peak picture elements and outputs an area, where these picture elements exist with a prescribed density or more, as a dot area. A bar-shaped area detection part 3 detects bar-shaped patterns such as '¦', '-'and '+' by performing pattern matching. When the bar-shaped pattern is detected, a general judge part 4 judges the noticed picture element as a character area regardless of the other detected results.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image area separation device for separating character areas with high accuracy.

[0002]

2. Description of the Related Art Conventionally, an image is separated into image areas and adaptively
In the case of performing the binarization process, in general, "image area separation method of mixed image of characters / pictures (dots and photographs)" (IEICE Transactions Vol. J75-DI1 No. 1 pp39-
47, January 1992), the character area and the pattern area are separated, and for example, the character area is subjected to MTF correction, then binarized with a fixed threshold, and the pattern area is dithered. Alternatively, binarization processing by the error diffusion method is performed to achieve both character resolution and pattern gradation.

[0003]

By the way, the above-mentioned image area separation method has a drawback that a grid-like pattern which should be originally determined as a character area is erroneously determined as a picture area. That is, as shown in FIG. 19, the “+” portion or the central portion of the “□” of the grid pattern is defined as the peak pixel of the halftone dot (pixel whose density level of the central pixel is higher or lower than that of the surrounding pixels). It is erroneously detected, and the whole area of the lattice pattern is erroneously determined to be a halftone dot area having a high peak pixel density, and the processing for the pattern is performed.

Further, for the image shown in FIG. 20 in which lines are densely arranged (depending on the SN of the scanner), the same processing as described above is performed. In other words, in the area where the lines are dense, if one line cannot be read at a uniform density level,
The ridge portion of the line shown in and of FIG. 20 may be erroneously detected as a peak pixel. by this,
The entire area of the pattern in which such lines are densely erroneously determined as a halftone dot area having a high peak pixel density, and processing for a pattern is performed.

Further, the present applicant has previously proposed an image area separation device that combines edge detection and white background detection (Japanese Patent Application No. 3-35752). In this image area separation device,
If the pixel of interest is an edge area and a white background exists in the vicinity, the pixel of interest is determined to be a character area.

A white background area detection circuit is provided in this apparatus, and a white background detection pattern shown in FIG. 21C is prepared in the circuit. Then, depending on the size of the white background pattern to be prepared (the larger the size of this pattern), the above-mentioned "lattice pattern" (see FIG. 21).
(A)) and "dense lines" (Fig. 21 (b)),
Since it is determined that there is no white background, there is a problem that it is mistakenly processed as a pattern as a result.

It is an object of the present invention to provide an image area separation device in which the accuracy of separation is improved by determining a dense portion of a grid pattern and lines as a character area.

[0008]

In order to achieve the above object, in the invention described in claim 1, means for inputting image information, means for converting the image information into a digital multi-valued signal, and a pixel of interest are provided. A means for detecting at least two or more different area information from a digital multi-valued signal or a binary signal in the vicinity thereof and a means for judging whether or not the pixel of interest is a character area based on the detection result are provided. In the image area separating apparatus, when the detecting unit detects a specific feature amount, the determining unit determines that it is a character area regardless of other feature amounts detected by the detecting unit.

According to a second aspect of the present invention, the specific feature amount is a local shape using pattern matching.

According to a third aspect of the invention, the local shape is a rod-shaped pattern.

According to a fourth aspect of the invention, the local shape is a bar pattern in the vertical and horizontal directions.

According to a fifth aspect of the invention, the local shape is a pattern in which rod-shaped patterns intersect in the vertical and horizontal directions, and the image area separating apparatus according to the second aspect.

According to the sixth aspect of the invention, the other feature amount is halftone dot area information indicating whether or not the pixel of interest is a halftone dot area.

According to a seventh aspect of the present invention, the halftone dot area information is information that is detected based on the density of the pixels by detecting the pixels corresponding to the peaks of the halftone dots.

According to the eighth aspect of the present invention, the other feature is white background area information indicating whether or not the vicinity of the pixel of interest is a white background area.

In the invention according to claim 9, the white background area information is information indicating whether or not a group of white pixels equal to or smaller than a predetermined threshold is detected and whether or not there is a group of white pixels in the vicinity of the pixel of interest. It is characterized by that.

[0017]

The edge area detection unit detects a continuous portion of white pixels and a continuous portion of black pixels and outputs a portion where both are present as an edge area, and the halftone area detection unit detects a peak pixel. , A region having peak pixels of a predetermined density or more is output as a halftone dot region, and the rod-shaped region detection unit detects a rod-shaped pattern by pattern matching. The comprehensive determination unit, when the rod-shaped area detection unit detects the rod-shaped pattern,
Regardless of the detection results of the edge area detection unit and the halftone dot area detection unit, the pixel of interest is determined to be the character area. In this way, when information with high accuracy that it is a character area is obtained,
Since the information is preferentially used and other information is not used, the character area can be separated with high accuracy.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be specifically described below with reference to the drawings. FIG. 1 is a block diagram of the first embodiment of the present invention. The image area separation device of the present invention is mainly composed of four processing blocks. That is, the edge area detection unit 1, the halftone dot area detection unit 2, the rod-shaped area detection unit 3,
Then, it is composed of the comprehensive judgment unit 4.

Hereinafter, each component will be described in detail. <Edge Area Detection Unit> For this detection unit, for example, the technique described in the above-mentioned paper (see FIG. 10 on p42) is used. FIG. 2 is a block diagram of the edge area detection unit. The ternarization circuit 11 performs edge enhancement on the input image data, and then ternarizes it using two types of fixed threshold values. The black pixel continuity detecting unit 12 and the white pixel continuity detecting unit 13 detect the regions where the black pixels and the white pixels after the ternarization are continuous by pattern matching.

FIG. 3 shows a pattern for detecting continuity of black pixels, and FIG. 4 shows a pattern for detecting continuity of white pixels. The continuous portion of white pixels and the continuous portion of black pixels matching these patterns are detected, and the AND circuit 14 outputs the portion where both are present as an edge area.

<Dot Area Detection Unit> This detection unit is also described in, for example, the above-mentioned paper (see FIG. 6 on p41).
Technology of. FIG. 5 is a block diagram of the halftone dot area detection unit. The peak pixel detector 21 detects peak pixels as shown in FIG. Here, the peak pixel is
The density level L of the central pixel is higher or lower than that of all the surrounding pixels, and the density levels a and b of the diagonal pixels that are diagonally across L and the central pixel are 4 | When xL−a−b |> TH (fixed threshold value), the central pixel is set as the peak pixel.

The halftone dot area detection circuit 22 sets 4 × 4 pixels to 1
If there are two or more blocks including the peak pixel in the four blocks, the block of interest is determined as a halftone dot candidate area, and the halftone dot area correction circuit 23 selects nine blocks centered on the target block. If 4 or more blocks are halftone dot candidates (that is, a halftone dot area exists with a certain size), the target block is the halftone dot area, and if not, the target block is the non-halftone dot area. As a correction process.

As described above, the "lattice pattern" and the "dense lines" are often matched with such peak pixels.

<Bar-Shaped Area Detection Unit> Before describing the details of the bar-shaped area detection unit, experimental facts will be described first. When a document image is read by a 400-dpi, 6-bit scanner (for example, a monochrome luminance signal and a color green signal), FIGS. 7 (a), 7 (b), 7 (c),
A rod-shaped pattern (rod-shaped pattern having a width of 1 pixel) as shown in (d) rarely appears in a picture (photograph, halftone dot).

Further, if the pattern is limited to the vertical direction and the horizontal direction, it is almost impossible to match the pattern. What is particularly problematic in the bar pattern detection of FIG. 7 is actually a halftone dot original. However, since the halftone dots generally have a certain screen angle, the problem of frequent matching with the bar pattern is greatly reduced if the halftone dots are limited to the vertical and horizontal directions.

Further, for the same reason, if a "+" character pattern as shown in FIG. 8 is prepared, only the grid pattern of the line drawing can be detected, and it is prevented from being erroneously detected in the pattern. In the example of FIG. 7, a bar pattern having a length of 5 pixels is used. However, by making this a bar pattern having a longer length such as 7 pixels or 9 pixels, a pattern is matched with such a pattern. The number of points is reduced (however, the cost increases by preparing a long bar pattern such as 7 pixels or 9 pixels).

By the way, the vertical and horizontal bar patterns of FIG. 7 and FIG.
Photographs and halftone dot originals that match the "+" character pattern of are, for example, "signs written on a white sign" in the design.
It is a vertical and horizontal bar on a white background, and is therefore the place where characters should be originally processed.

In view of the above background, in the present invention, the regions particularly belonging to the bar patterns in the vertical and horizontal directions as shown in FIGS. 7 and 8 (that is, the target pixel in FIGS. 7 and 8 and its neighboring pixels). Is based on the idea that it is determined as a character area regardless of other information, for example, the detection result of a halftone dot area.

Now, FIG. 9 is a block diagram of the rod-shaped area detection unit. The binarization circuit 31 binarizes the signal from the scanner with a fixed threshold value. As a pre-treatment, MTF
It is also effective to apply a correction filter (for example, the filter shown in FIG. 11 of the above-mentioned paper). Further, it is possible to expect the effect of reducing the erroneous determination in the pattern (especially the photograph) by detecting the black pixel and the white pixel by performing the ternarization with two different thresholds.

In the pattern matching circuit 32, as shown in FIG.
Rod-shaped pixels are detected using the "|" pattern of (a), the "-" pattern of (b), the "+" pattern of FIG. 8, and the patterns shown in FIGS. 10 (a) to 10 (d). Also, these patterns are patterns that detect fine lines on a white background, but by reversing the black and white of the prepared pattern,
White characters in a black solid can also be detected. When the patterns match, the pixel of interest is output as a rod-shaped pixel.

In the expansion circuit 33, the pixel area subjected to the pattern matching as described above is converted into 5 × as shown in FIG.
Inflate with a mask of 5. That is, 1 in a 5x5 mask
If even individual rod-shaped pixels exist, the pixel of interest is made a rod-shaped pixel.

The reason why such expansion processing is performed is that the pixel position matched by the pattern matching circuit 32 is the spine (ridge) portion of the character, and the portion to be actually determined to be the character area is the character edge. The expansion process as described above is required.

<Comprehensive Judgment Unit> In the comprehensive judgment unit 4, the following three detection results are obtained for each pixel: edge (1) / non-edge (0) in the edge area detection unit, halftone dot (1) / in the halftone area detection unit Non-halftone dot (0) Based on the rod-like (1) / non-rod-like (0) in the rod-like area detection unit, the attribute (character area / pattern area) of the pixel of interest is determined.

FIG. 12 shows a judgment table. The feature of this table is that if the detection result of the rod-shaped area detection unit is rod-shaped (the output signal is 1), the pixel of interest is determined to be the character area regardless of other detection results.

As described above, according to the present invention, when the character area / pattern area is determined, the highly accurate information from the rod-shaped area detecting portion is used with priority, so that the character area is separated with high accuracy. be able to.

FIG. 13 is a block diagram showing another embodiment of the present invention. In the image area separating apparatus of this embodiment, the halftone dot area detecting unit of the above-described embodiment is replaced with the white background area detecting unit 5, and the edge area detecting unit and the rod-like area detecting unit are the same as those described above. Therefore, the white background area detection unit 5 and the comprehensive determination unit 4 will be described.

As this detection unit, for example, the one previously proposed by the present applicant (Japanese Patent Application No. 3-35752) is used. FIG. 14 shows the configuration of the first embodiment of the white background area detection unit 5. The white background area refers to an area in which a block of white pixels having a predetermined size (1 × 5 or 5 × 1) exists near the target pixel.

After the input image signal is sharpened by the MTF correction circuit 51, the binarization circuit 52 binarizes the image signal into white / non-white by a predetermined threshold value and inputs it to the white pixel block pattern matching circuit 53. . FIG. 15 shows a pattern of white pixel blocks in which 1 × 5 or 5 × 1 are all white pixels.
The white pixel block pattern matching circuit 53 detects a block of white pixels near the target pixel by pattern matching,
When the pattern shown in FIG. 15 is matched, the pixel of interest is set as the active pixel. From the output of the white pixel block pattern matching circuit 53, the expansion circuit 54 counts the number of active pixels in, for example, a 5 × 5 block centered on the pixel of interest, and if there is at least one active pixel in the block, The entire block is judged and output as a white background area.

FIG. 16 shows a second embodiment of the white background area detecting section. In this embodiment, a correction circuit is provided. Other configurations are similar to those of the first embodiment. Input image signal is M
After the sharpening processing in the TF correction unit, the binarization circuit binarizes the image signal into white / non-white according to a predetermined threshold value, and inputs it to the white pixel block pattern matching circuit. The white pixel block pattern matching circuit detects a block of white pixels in the vicinity of the pixel of interest by pattern matching, and when matching the pattern shown in FIG. 15, sets the pixel of interest as a temporary active pixel.

After detecting the cluster of white pixels, in the correction circuit,
As shown in FIG. 17, when there are clusters A and B of white pixels at a distance of L pixels left and right from the target pixel in the main scanning direction, the target pixel is output as a true active pixel. To do. The expansion circuit counts the number of active pixels in, for example, a 5 × 5 block centered on the pixel of interest, and if even one active pixel exists in the block, the entire block is determined and output as a white background area.

As described above, the definition of the white background (lump of white pixels) is explained as 5 × 1 or 1 × 5 in the above example, but in reality, a mask of a larger size must be prepared. In some cases, "lattice pattern" and "dense lines" will match.

<Comprehensive Judgment Unit> In the comprehensive judgment unit 4, the following three detection results are obtained for each pixel: Edge (1) / non-edge (0) in the edge area detection unit White background (1) / non-white background in the white background area detection unit (0) Based on the rod-shaped (1) / non-rod-shaped (0) in the rod-shaped region detection unit, the attribute (character region / pattern region) of the pixel of interest is determined.

FIG. 18 shows a judgment table. Similar to the previous table, the feature of this table is that if the detection result of the rod-shaped area detection unit is rod-shaped (the output signal is 1), the pixel of interest is determined to be the character area regardless of other detection results. It is a point.

[0044]

As described above, according to the first aspect of the present invention, when the information indicating the character area is surely obtained, the information is prioritized and other information is not used. Therefore, accurate separation can be performed. In addition, since they are separated by local processing, they can be realized by a small-scale circuit configuration.

According to the second to fifth aspects of the invention, the information that the character area is surely obtained by the pattern matching using the pattern such as the rod-shaped pattern and the "+" character pattern, and the information is prioritized. Since no other information is used, highly accurate character area separation is possible.

According to the sixth and seventh aspects of the present invention, when the information indicating that the character area is surely obtained, since the information of the halftone dot separation is not used, the part which cannot be separated by the halftone dot area detecting unit. (That is, the grid-shaped lines and the densely packed lines) can be covered, and accurate separation can be achieved.

According to the eighth and ninth aspects of the present invention, when the information indicating that the character area is surely obtained, the information on the white background separation is not used. (That is, the grid-shaped lines and the densely packed lines) can be covered, and accurate separation can be achieved.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of a first embodiment of the present invention.

FIG. 2 is a block configuration diagram of an edge area detection unit.

FIG. 3 shows a pattern for detecting continuity of black pixels.

FIG. 4 shows a pattern for detecting continuity of white pixels.

FIG. 5 is a block diagram of a halftone dot area detection unit.

FIG. 6 is a diagram illustrating peak pixels.

7 (a), (b), (c) and (d) are examples of rod-shaped patterns.

FIG. 8 is an example of a “+” pattern.

FIG. 9 is a block configuration diagram of a rod-shaped area detection unit.

10 (a), (b), (c) and (d) are other examples of rod-shaped patterns.

FIG. 11 is a 5 × 5 dilation mask.

FIG. 12 is a comprehensive determination table according to the first embodiment.

FIG. 13 is a block diagram showing another embodiment of the present invention.

FIG. 14 shows a configuration of a first embodiment of a white background area detection unit.

FIG. 15 shows a pattern of white pixel blocks in which 1 × 5 or 5 × 1 are all white pixels.

FIG. 16 is a second example of a white background area detection unit.

FIG. 17 is a diagram for detecting a cluster of white pixels at a distance of L pixels to the left and right from the target pixel.

FIG. 18 is a comprehensive determination table according to the second embodiment.

FIG. 19 is an example of a grid pattern that is erroneously determined as a halftone dot area.

FIG. 20 is an example of a dense pattern of lines erroneously determined as a halftone dot area.

21A is an example of a grid-like pattern that is erroneously determined to have no white background, FIG. 21B is an example of a dense pattern of lines that is erroneously determined to have no white background, and FIG. It is a white background pattern.

[Explanation of symbols]

 1 Edge Area Detection Section 2 Halftone Area Detection Section 3 Rod Area Detection Section 4 Total Judgment Section 5 White Area Detection Section

Claims (9)

[Claims] 1. A means for inputting image information, a means for converting the image information into a digital multi-valued signal, and at least two or more different area information from a pixel of interest and a digital multi-valued signal or a binary signal in the vicinity thereof. In an image area separating device comprising a means for detecting and a means for determining whether or not the pixel of interest is a character area based on the detection result, when the detecting means detects a specific feature amount, An image area separation device, wherein the determination means determines that the area is a character area regardless of the other characteristic amounts detected by the detection means. 2. The image area separation device according to claim 1, wherein the specific feature amount is a local shape using pattern matching. 3. The image area separation device according to claim 2, wherein the local shape is a rod-shaped pattern. 4. The image area separation device according to claim 2, wherein the local shape is a bar-shaped pattern in a vertical direction and a horizontal direction. 5. The image area separation device according to claim 2, wherein the local shape is a pattern in which rod-shaped patterns intersect in the vertical and horizontal directions. 6. The image area separation device according to claim 1, wherein the other feature amount is halftone dot area information indicating whether or not the pixel of interest is a halftone dot area. 7. The image area according to claim 6, wherein the halftone dot area information is information determined by detecting a pixel corresponding to a peak of a halftone dot and determining a density of the pixel. Separation device. 8. The image area separation device according to claim 1, wherein the other feature is white background area information indicating whether or not the vicinity of the pixel of interest is a white background area. 9. The white background area information is information indicating whether or not a group of white pixels equal to or smaller than a predetermined threshold is detected and whether or not there is a group of white pixels near a target pixel. Item 8. The image area separation device according to item 8.