JP4846681B2 - Image processing apparatus, image processing method, and computer-readable storage medium storing image processing program - Google Patents

Description

  The present invention provides a monochrome multifunction device, a color multifunction device, or the like that performs character identification processing of an image area on an image read by a scanner or an image acquired via a network and processes the image according to the result. The present invention relates to an image processing apparatus, an image processing method, and a computer-readable storage medium storing an image processing program.

In an image processing apparatus such as a monochrome multifunction peripheral or a color multifunction peripheral, a demand for higher image quality has been increased with an increase in resolution of image processing technology.
In general image processing in such an image processing apparatus, various processing such as scanner γ correction processing, scaling processing, filter processing, and printer γ processing is performed. In particular, the filtering process is a process for correcting the sharpness of an image signal input from a scanner unit or the like, and determines the sharpness (crispness) of characters and the granularity (roughness) of a pattern in an output image. This is an important process.

If the emphasis is enhanced by the filter process, the sharpness of the character is improved, but the graininess of the pattern is deteriorated. If the emphasis is on the smoothing process by the filter process, the graininess of the pattern is improved, but the sharpness of the character is decreased. Run short.
Therefore, in the conventional image forming apparatus, the character is identified from the image area before performing the filtering process, and the image area is separated into the character area and the non-character area, and the optimum filtering process according to the attribute of the image is performed. I do.
However, when there are halftone dots in the image area and “characters on halftone dots” are identified (see “characters on halftone dots” in FIGS. 85 (b) and 86 (b)), In some cases, a halftone dot may be erroneously identified, and a smoothing process or the like performed on a pattern on a character on the halftone dot is executed, which makes it difficult to read the character on the halftone dot.

Therefore, Patent Document 1 below is a method of separating a character part from a multi-tone image in which a character part and a picture part are mixed, and each pixel is classified into a black pixel and a white pixel, A connected black pixel and a connected white pixel in the area are detected, and when a certain number or more of the connected black pixels and white pixels exist, it is determined whether the target pixel is a character line drawing area or a halftone dot area. Technology is disclosed.
In Patent Document 2 below, input digital image data is binarized using an average density of m × n blocks as a threshold, and the continuity of the binary signal is detected, so that the pixel of interest is a character line drawing region or a halftone dot. A technique for determining whether or not an area is present is disclosed.
In Patent Document 3 below, the smoothness of an edge is determined by paying attention to the local unevenness of the edge in the image area, and the character edge on the halftone dot is separated from the edge in the halftone dot pattern from the smoothness of the edge. An image processing apparatus for identification is disclosed.

JP-A-2-292957 JP-A-1-227573 Japanese Patent Application No. 2007-60529 (unpublished patent document)

However, in these patent documents, “characteristic part of a character having a partially small shape” on a halftone dot (characteristic part of a character having a partially small shape is, for example, a cloud point, a punctuation mark, or a hiragana character. The last corner part of “O”, the part where the line is isolated and has a point shape like the point part of “・”, “■”, and the small round shape like a donut such as a punctuation point, a semi-turbid point, etc. If there is a part that exists in the image, the number of pixels of the character will be less than the number of pixels necessary for the identification process, which causes the problem of making it difficult to identify the character. 47 (a)).
In addition, when there are unclear and small characters (for example, characters whose resolution is lower than 600 dpi (dot per inch) and the number of fonts is lower than 6 pt (point)) on the halftone dot, the number of pixels of the character is identified. Therefore, there is a problem that it is difficult to identify characters.
Furthermore, when there is a character having a shape in which the edge direction on the halftone dot is frequently changed, there arises a problem that it becomes difficult to identify the cross of the character part or the intersection of the T character (the T character part, and (See FIG. 87 (b) for the cross).

Therefore, the present invention provides an image processing apparatus that cuts a character edge on a halftone dot from an edge in a halftone dot pattern, enables identification of a blurred and small character, and performs character identification with high accuracy without omission. This is the first purpose.
A second object is to make it possible to identify isolated short lines such as cloud marks and reading marks, and to perform character identification with high accuracy without omission.
A third object is to make it possible to identify small round shapes such as semi-turbids and punctuation marks, and to perform character identification with high accuracy without omission.
A fourth object is to make it possible to identify a portion where character portions intersect with a cross, and to perform character identification with high accuracy without omission.
It is a fifth object of the present invention to make it possible to identify a portion where a character portion intersects a T-shape and to perform character identification with high accuracy without omission.
A sixth object of the present invention is to provide an image processing method in which a character edge on a halftone dot is separated from an edge in a halftone dot pattern, enables identification of unclear and small characters, and performs character identification with high accuracy without omission. Objective.
A computer-readable image processing program that cuts character edges on halftone dots from the edges in the halftone dot pattern, enables identification of unclear and small characters, and performs character identification with high accuracy without loss A seventh object is to provide a storage medium.

According to the first aspect of the present invention, image data receiving means for receiving image data , first pattern matching processing, horizontal pixel count processing, and vertical direction for the image data received by the image data receiving means Image data received by the image data receiving means, the basic identification means for performing pixel count processing and identifying the character characteristics of the characters on the halftone dot in which the number of pixels in which black pixels or white pixels continue is equal to or greater than the predetermined number of pixels On the other hand, continuous white pixel extraction processing, extraction processing of an area surrounded by white pixels, black pixel block counting processing, isolated point extraction processing, checker pattern extraction processing, second pattern matching processing, and third pattern performs a matching process, different from the character features of the identified dot on characters by the base identifying means, black pixels, or number of pixels white pixels are consecutive fully a predetermined number of pixels The character features on no halftone characters, a supplemental identification means for identifying as a character feature on halftone characters, the basic identification means identified dot on character features, and identified by the supplementary identification means Image attribute determining means for separating the image data received by the image data receiving means into a character area and a non-character area on the basis of the character characteristics of the characters on the halftone dot. Achieve.
The invention according to claim 2 is characterized in that, in the invention according to claim 1, the basic identification means identifies a character feature of the character on the halftone dot as an edge portion.
According to a third aspect of the present invention, in the first or second aspect of the present invention, the basic identification unit is configured such that the number of pixels in which black pixels or white pixels are continuous is equal to or greater than a predetermined number of pixels in a certain direction. In this case, the character feature is identified as a character feature of a character on a halftone dot.
According to a fourth aspect of the present invention, in the first, second, or third aspect of the invention, the basic identifying means identifies the character feature of the character on the halftone dot as a ridge pixel. .
According to a fifth aspect of the present invention, in the first, second, third, or fourth aspect, the supplementary identifying means identifies the character feature of the character on the halftone dot as an isolated part. Thus, the second object is achieved.
According to a sixth aspect of the present invention, in the first, second, third, fourth, or fifth aspect, the supplementary identifying means rounds the character characteristics of the halftone characters. The third object is achieved by identifying the portion forming the shape.
According to a seventh aspect of the present invention, in the first, second, third, fourth, fifth, or sixth aspect, the supplementary identifying means is a character of the halftone dot character. The fourth object is achieved by identifying the feature as a portion where the line intersects the cross.
In the invention according to claim 8, in the invention according to claim 1, claim 2, claim 3, claim 4, claim 5, claim 6 or claim 7, the supplementary identification means is the halftone dot. The fifth object is achieved by identifying the character feature of the upper character as a portion where the line intersects the T-shape. According to the ninth aspect of the present invention, in the image processing apparatus provided with the image data receiving means for receiving the image data , the first pattern matching process, the number of pixels in the horizontal direction for the image data received by the image data receiving means. A first step of performing a count process and a vertical pixel count process, and identifying a character feature of a character on a halftone dot in which the number of pixels in which black pixels or white pixels are continuous is equal to or greater than a predetermined number of pixels; Continuous white pixel extraction processing, extraction processing of a region surrounded by white pixels, black pixel block counting processing, isolated point extraction processing, checker pattern extraction processing, second pattern matching for image data received by the receiving means process, and the third performs pattern matching process, different from the character features of the first on halftone identified in step character, black pixels or white pixels are The character features of a halftone dot on a character which the number of pixels is less than the predetermined number of pixels to be continued, on the halftone characters and the second identifying as a character feature, the first on halftone identified in step character A third step of separating the image data received by the image data receiving means into a character region and a non-character region based on the character feature and the character feature of the halftone character identified in the second step; The sixth object is achieved.
According to the tenth aspect of the present invention, an image data receiving function for receiving image data , and a first pattern matching process, a horizontal pixel count process, and a vertical direction for the image data received by the image data receiving function Image data received by the basic identification function that performs the pixel count processing and identifies the character characteristics of the characters on the halftone dot in which the number of black pixels or continuous white pixels is equal to or greater than the predetermined number of pixels, and the image data receiving function On the other hand, continuous white pixel extraction processing, extraction processing of an area surrounded by white pixels, black pixel block counting processing, isolated point extraction processing, checker pattern extraction processing, second pattern matching processing, and third pattern It performs a matching process, the different character features of the identified dot on characters in the basic identification functions, black pixels or the number of pixels of white pixels are continuous within a predetermined number of pixels The character features Tanai dot on characters, identified by supplemental identification features and the basic identification features in the identified dot on character features, and the supplementary identification function of identifying as a character feature on halftone characters An image attribute determination function for separating the image data received by the image data reception function into a character area and a non-character area based on the character characteristics of the characters on the halftone dots. To achieve.

In the inventions according to claims 1 to 4, claim 9, and claim 10, the character edge on the halftone dot is separated from the edge in the halftone dot pattern, and an unclear and small character can be identified. Character identification can be performed with high accuracy without omission.
According to the fifth aspect of the present invention, it is possible to identify isolated short lines such as muddy spots and reading marks, and it is possible to identify characters with high accuracy without omission.
According to the sixth aspect of the present invention, it is possible to identify small round shapes such as a semi-turbid point and a phrase, and it is possible to identify characters with high accuracy without omission.
According to the seventh aspect of the present invention, it is possible to identify the portion where the character portion intersects the cross, and the character can be identified with high accuracy without omission.
According to the eighth aspect of the present invention, it is possible to identify the portion where the character portion intersects the T-shape, and the character can be identified with high accuracy without omission.

DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments for carrying out the invention will be described in detail with reference to the drawings.
[Details of First Embodiment]
FIG. 1 is a block diagram showing the configuration of the first exemplary embodiment of the present invention.
The scanner 10 reads an original and receives image data composed of RGB signals, and the scanner γ correction unit 11 corrects the γ characteristics of the scanner by converting attributes from a linear reflectance signal to a linear density signal.
In this attribute conversion, a density linear signal effective for smoothing is converted into a density linear signal effective for edge enhancement. The scanner color correction unit 12 converts the RGB signal depending on the scanner characteristics into a device-independent R′G′B ′ signal that does not depend on the scanner.
The image attribute determination means 20 extracts a character area from the image data, and extracts characters such as white ground characters and halftone dots as character areas.
The edge amount extraction unit 30 extracts an edge amount that is a signal representing the likelihood of an edge in the image from the R′G′B ′ signal.
Based on the results of the image attribute determination unit 20 and the edge amount extraction unit 30, the filter processing unit 13 performs a smoothing process that suppresses the undulation of the halftone part and suppresses moire, and an edge enhancement process that increases the sharpness of the character part. Do.
The printer color correction means 14 performs conversion from a device-independent R′G′B ′ signal to a C′M′Y ′ signal depending on printer characteristics.
The edge amount extraction means 40 extracts an edge amount that is a signal indicating the degree of edge (edge-likeness) in the image from the C′M′Y ′ signal. The UCR / black generation unit 15 generates a K signal according to the C′M′Y ′ data based on the results of the image attribute determination unit 20 and the edge amount extraction unit 40 (black generation), and C′M′Y ′. Under color removal (UCR) is performed by subtracting the amount corresponding to K from K. Note that CMYK at this time corresponds to each color of the printer.
The printer γ correction unit 16 performs printer γ correction on the signal after undercolor removal by the UCR / black generation unit 15, and further performs pseudo gradation processing by the pseudo gradation processing unit 17. An image is output on a recording medium.
More specifically, the printer γ correction means 16 performs conversion processing using a density conversion table in accordance with the density characteristics of the printer, and the pseudo gradation processing means 17 performs pseudo halftone processing such as dither and error diffusion. .
In the present embodiment, image data is received by reading a document from the scanner 10, but it is also possible to receive image data from communication with the Internet (not shown) or a storage medium.

FIG. 2 is a block diagram showing the configuration of the image attribute determination means 20. The RGB → CMY conversion 201 converts the signal into a cmy signal corresponding to the process color of the document. After the binarizing means 202, 212, and 222 for each of the cmy signals, the basic identification means 507, 508, and 509 are supplemented. Identification means 501, 502, and 503 are performed, and an OR operation is performed on the identification result identified from each cmy signal in OR208. If the signal is identified as a character by any one of the signals Cmy, the identification result of the image attribute determination means 20 is “character”.
The RGB → CMY conversion 201 converts the R′G′B ′ signal into a cmy signal corresponding to the process color of the document by the following equation.
c = a0 + a1 × R + a2 × G + a3 × B
m = b0 + b1 × R + b2 × G + b3 × B
y = c0 + c1 * R + c2 * G + c3 * B (Formula 1)
In Expression 1, a0 to a3, b0 to b3, and c0 to c3 are parameters set in advance based on the relationship between the RGB read value and the color of the color patch when the process color patch is read by the scanner.

FIG. 3 is a block diagram showing the configuration of the binarization means 202, 212, 222. The peak pixel detection means 231 and 232 detect whether or not the target pixel is an extreme point indicating a peak of density change from the density relationship with surrounding pixels.
For example, when the density level of the central pixel is higher than all other density levels in a block composed of M × M pixels, the following (Expression 2) or (Expression 3) shown below is performed. Determine if it is a pole.
The peak pixel is detected by the peak pixel detecting means 231 according to (Equation 2) shown below, and the peak pixel is detected by the peak pixel detecting means 232 according to (Equation 3) shown below. is there.
(1) When M = 3 (FIG. 4A),
(2m0-m1-m8) ≧ ΔmTH and
(2m0−m2−m7) ≧ ΔmTH and
(2m0−m3−m6) ≧ ΔmTH and
(2m0-m4-m5) ≧ ΔmTH (Formula 2)
(2) In the case of M = 5 (FIG. 4B),
(2m0−m3−m22) ≧ ΔmTH and
(2m0-m8-m17) ≧ ΔmTH and
(2m0-m1-m24) ≧ ΔmTH and
(2m0−m7−m18) ≧ ΔmTH (Formula 3)

The peak pixel detection means 231 and 232 detect the central pixel as a peak when the density difference between the average value of the two pixel levels and the central pixel located symmetrically with respect to the central pixel is larger than the threshold value ΔmTH. The OR 233 determines that the pixel is a peak pixel if any one of the peak pixel detector 231 and the peak pixel detector 232 is detected as a peak pixel.
The threshold value setting means 234 has the highest pixel value (highest density) from the peak pixel and the pixel value (c or m or y value) within the 9 × 9 pixel centering on the target pixel. Are set as threshold = peak pixel maximum value + α. α is a parameter for giving a margin to the threshold value, and is set to a value of α = about 10 to 30 when the CMY signal is 8 bits each.
The white pixel / black pixel determination unit 235 determines a black pixel if the image data is equal to or greater than a set threshold value, and determines a white pixel if the image data is less than the set threshold value.
As a result, the halftone dot character is determined to be a white pixel, and the character portion has a higher density than the halftone dot (with the exception of white dots on the dot), so it is determined as a black pixel. Is done.

FIG. 5 is a block diagram showing the configuration of the edge smoothness determination means 203, 213, and 223. For 3 × 3 size binary image data, each pattern matching means performs the following pattern matching (1) to (4).
(1) The P1 pattern matching unit 241 performs pattern matching indicated by P1 in FIG.
(2) The P2 pattern matching unit 242 performs pattern matching indicated by P2 in FIG.
(3) The P3 pattern matching unit 243 performs pattern matching indicated by P3 in FIG.
(4) The P4 pattern matching means 244 performs pattern matching indicated by P4 in FIG.
As a result of the above (1) to (4), each pattern matching means outputs “1” if they match, and “0” if they do not match.

In the horizontal direction pixel number counting means 245 and 246 and the vertical direction pixel number counting means 247 and 248, the number of pixels in the horizontal direction or the vertical direction is counted for each output value, and the area in the four directions as viewed from the target pixel. Extract vertical edges or horizontal edges.
At this time, it is extracted as eight binary signals of upper left vertical edge / lower left vertical edge / upper right vertical edge / lower right vertical edge / upper left horizontal edge / upper right horizontal edge / lower left horizontal edge / lower right horizontal edge.

In FIG. 7B, the reference area is divided into four equal parts with the pixel of interest at the center, and when the left edge of the vertical line (rising edge) is in the upper left area, it is extracted as the upper left vertical edge, and the left edge of the vertical line is extracted in the lower left area It is the figure which showed the example extracted as a lower left vertical edge when there exists (rising edge).
FIGS. 7A and 7C are diagrams showing an example in which the upper left vertical edge and the lower left vertical edge are extracted when there is an inclination of 45 ° to 135 °.
In FIG. 8B, when there is a vertical right edge (falling edge) in the upper right area, it is extracted as an upper right vertical edge, and when there is a vertical line right edge (falling edge) in the lower right area, the lower right vertical It is the figure which showed the example extracted as an edge.
FIGS. 8A and 8C are diagrams showing an example in which the upper right vertical edge and the lower right vertical edge are extracted when there is an inclination of 45 ° to 135 °.
FIG. 9B divides the reference area into four equal parts with the pixel of interest at the center, and when the upper left area has a horizontal upper edge (rising edge), the upper left horizontal edge, and the upper right area has a horizontal line upper edge (rising edge). It is the figure which showed the example extracted as an upper right horizontal edge at a certain time.
FIGS. 9A and 9C are diagrams illustrating an example in which the upper left horizontal edge and the upper right horizontal edge are extracted even when there is an inclination of −45 ° to 45 °.
In FIG. 10B, when the lower left edge (falling edge) of the horizontal line is in the lower left area, it is extracted as the lower left horizontal edge, and when the lower edge (falling edge) of the horizontal line is in the lower right area, It is the figure which showed the example extracted as a lower horizontal edge.
FIGS. 10A and 10C are diagrams showing an example in which the lower left horizontal edge and the lower right horizontal edge are extracted when there is an inclination of −45 ° to 45 °.
7 to 10 exemplify a relatively thin line in which a pair of edges exist in the reference area, but a pair of edges exist in the reference area as shown in FIG. Similarly, an edge such as a thick line that is not extracted is extracted as an upper left vertical edge and a lower left vertical edge.
As shown in FIG. 11B, not only a straight line but also an upper left vertical edge and a lower left vertical edge are extracted at the edge of a smooth curve.

FIG. 12 is a diagram for explaining processing of the horizontal pixel number counting unit 245 and the horizontal pixel number counting unit 246.
An example will be described in which the pattern matching result of 13 pixels in the horizontal direction across the target pixel is as shown in (1).
The pattern matching result is multiplied by (2) the weight for counting the number of horizontal pixels corresponding to the pixel position.
The weight is zero at the pixel position of interest, and the absolute value increases as the distance increases, with one direction being negative and the opposite direction being positive.
That is, the result of (1) × (2) is obtained by determining how many pixels away from the pixel of interest the pixel matching the pattern is located in the left or right direction.
The position of the pixel closest to the target pixel among the pixels matched with the pattern (select the one with the smallest absolute value from the result of (1) × (2)) is output as the result of the horizontal pixel count.
However, (1) if the pattern matching result is all zero in the reference area, the pixel count result is the maximum value “7”.
FIG. 13 is a diagram for explaining processing of the vertical pixel number counting unit 247 and the vertical pixel number counting unit 248. A case where the pattern matching result of 13 pixels in the vertical direction across the target pixel is as shown in (1) will be described as an example.
The pattern matching result is multiplied by (2) a weight for counting the number of vertical pixels corresponding to the pixel position. This weight is zero at the target pixel position, and the absolute value increases as the distance from the pixel position increases, with one direction being negative and the opposite direction being positive.
That is, the result of (1) × (2) is an index that determines how many pixels away from the pixel of interest the pixel matching the pattern is located in the vertical direction.
Then, the position of the pixel closest to the target pixel among the pixels matching the pattern (select the one with the smallest absolute value from the result of (1) × (2)) is output as the result of counting the number of pixels in the vertical direction. To do.
However, if the pattern matching result is all zero in the reference area in (1), the maximum number “7” is output as the pixel count result.

FIG. 14 is a diagram for explaining reference regions and extraction conditions of the upper left vertical edge extraction unit 249 and the upper right vertical edge extraction unit 251. The horizontal pixel count results y1 to y7 are referred to the upper seven pixels including the target pixel (y7) in the vertical direction, and (1) vertical line extraction conditions and (2) diagonal line extraction conditions (smooth curve) 1 is output as an upper left vertical edge or an upper right vertical edge, and “0” is output when neither is satisfied.
The first condition for both (1) vertical line extraction condition and (2) diagonal line extraction condition is that the result of the horizontal pixel count of all seven pixels is smaller than “7”, that is, the horizontal direction is always That is, there is a pixel matched by pattern matching.
(1) The second condition of the vertical line extraction condition is that the difference between the maximum value and the minimum value of the horizontal pixel count is 1 or less.
(2) The second condition of the diagonal line extraction condition is that the horizontal pixel count results are arranged in descending order from top to bottom.
(1) The vertical line extraction condition is not only a vertical line with no noise, but also a vertical line with a 1 dot projection as shown in the example of a character image on a halftone dot in FIG. It is a condition to extract as a line, and the determination condition is slightly relaxed by an oblique line.

  FIG. 15 is a diagram for explaining reference regions and extraction conditions of the lower left vertical edge extraction unit 250 and the lower right vertical edge extraction unit 252. The results of the horizontal pixel count y7 to y13 are referred to the lower seven pixels including the target pixel (y7) in the vertical direction. The rest is the same as the upper left vertical edge extraction 249 and the upper right vertical edge extraction 251 of FIG.

FIG. 16 is a diagram for explaining reference regions and extraction conditions of the upper left horizontal edge extraction unit 253 and the lower left horizontal edge extraction unit 255. Refer to the left pixel including the pixel of interest (y7) in the horizontal direction from the result of the vertical pixel count t1 to t7,
If either (1) the horizontal line extraction condition or (2) the diagonal line extraction condition (including a smooth curve) is satisfied, “1” is output as the upper left horizontal edge or the lower left horizontal edge. "0" is output.
The first condition for both (1) the horizontal line extraction condition and (2) the diagonal line extraction condition is that the result of counting the number of vertical pixels in all seven pixels is smaller than “7”, that is, the pattern is always in the vertical direction. That is, there is a pixel matched by matching.
(1) The second condition of the horizontal line extraction condition is that the difference between the maximum value and the minimum value of the vertical direction pixel count is 1 or less.
(2) The second condition of the diagonal line extraction condition is that the results of the vertical pixel count are arranged in descending order from left to right.
(1) The horizontal line extraction condition is not only a horizontal line with no noise, but also a horizontal line with a 1-dot protrusion, so that it is extracted as a vertical line. A little loose.

FIG. 17 is a diagram for explaining reference regions and extraction conditions of the upper right horizontal edge extracting unit 254 and the lower right horizontal edge extracting unit 256.
As a result of counting the number of pixels in the vertical direction, the right seven pixels including the target pixel (t7) are referred to in the horizontal direction.
The rest is the same as the upper left horizontal edge extracting means 253 and the lower left horizontal edge extracting means 255 of FIG.
Among the extraction conditions for each edge described above, the one described as the second condition corresponds to the condition for determining the smoothness of the edge, and determines the presence or absence of local unevenness in the reference region.

  FIG. 18 is a block diagram showing the configuration of the continuous edge detection means 204, 214, 224. As shown in FIG. 18, the continuous edge detecting means 204, 214, 224 are configured by AND operation (AND) and OR operation (OR). For example, as shown in FIG. In the case of the lower left vertical edge, it is detected as “continuous edge”. In this way, continuous edges are detected when smooth lines are continuous, and mainly the edges of thick characters and lines are detected.

FIG. 19 is a block diagram showing the configuration of the pair edge detection means 205, 215, 225. As shown in FIG. 19, the pair edge detection means 205, 215, and 225 are configured by a logical product operation (AND) and a logical sum operation (OR). For example, as shown in FIG. An edge and an upper right vertical edge, or a lower left vertical edge and a lower right vertical edge are detected as “pair edges”.
Thus, mainly thin characters and lines are detected as pair edges. As shown in FIG. 22B, edges near the end points are also detected as pair edges.

FIG. 20 is a block diagram showing the configuration of the intersection edge detection means 206, 216, 226. As shown in FIG. 20, the intersection edge detection means 206, 216, and 226 are composed of a logical product operation (AND) and a logical sum operation (OR). For example, as shown in FIG. In the case of the lower left vertical edge, it is detected as an “intersection edge”. In this way, mainly the intersections of the lines constituting the characters, the end points of the bold characters, and the corners of the broken lines are detected.
FIG. 24 is a block diagram showing the configuration of the character / picture pattern identifying means 207, 217, 227. A bold character (bold line) edge and a thin character (thin line) edge are identified, respectively, and if it is a bold character edge or a fine character edge, it is identified as a character edge in OR295.
The bold edge extraction is performed by the correction unit 291 and the result is expanded by the expansion unit 292. For a bold edge, a continuous edge and an intersection edge should be extracted from continuous edges / pair edges / intersection edges.
The correcting means 291 outputs “1” if a continuous edge or intersection edge is detected in all the pixels in the 5 × 5 reference region, and “0” if there is a pixel that is neither a continuous edge nor an intersection edge. Is output.

The expansion unit 292 refers to the result of the correction unit 291 for 5 × 5 pixels, and if “1” exists, extracts it as a bold character edge. The extraction of the fine character edge is corrected by the correction 293, and the result is expanded by the expansion means 294.
In the fine character edge, the pair edge and the intersection edge are extracted from the continuous edge / pair edge / intersection edge.
The correction unit 293 outputs “1” if a pair edge or intersection edge is detected in all the pixels in the 5 × 5 reference region, and outputs “0” if a pixel that is neither a pair edge nor an intersection edge exists. To do. The expansion unit 294 refers to the result of the 5 × 5 pixel correction unit 293, and if “1” exists, extracts it as a fine character edge.

Next, specific processing of edge smoothness determination will be described with two specific examples in the case of a character edge on a halftone dot and the case of a halftone dot pattern edge.
(Specific example 1 of edge smoothness determination)
First, the edge smoothness determination in the case of a character edge on a halftone dot will be specifically described below.
FIG. 25 is a diagram showing an image after binarization of a character edge on a halftone dot.
FIG. 26 shows the result of the P1 pattern matching unit 241 and the result of the horizontal pixel count unit 245 for the image of FIG. 25, and there is a horizontal pixel count result = 7 for both the upper left vertical edge and the lower left vertical edge. Therefore, the extraction result is “0”.
FIG. 27 shows the result of the P2 pattern matching unit 242 and the result of the horizontal pixel count unit 246 for the image of FIG. 25, and there is a horizontal pixel count result = 7 for both the upper right vertical edge and the lower right vertical edge. Therefore, the extraction result is “0”.
FIG. 28 shows the result of the P3 pattern matching unit 243 and the result of the vertical pixel number counting unit 247 for the image of FIG. 25, and the vertical pixel number count result is less than 7 for both the upper left horizontal edge and the upper right horizontal edge. And since it is arranged in descending order from the left to the right, the extraction result is “1”.
FIG. 29 shows the result of the P4 pattern matching unit 244 and the result of the vertical pixel number counting unit 248 for the image of FIG. 25, and the vertical pixel number count result is less than 7 for both the lower left horizontal edge and the lower right horizontal edge. And since it is arranged in descending order from the left to the right, the extraction result is “1”.

As a result of the edge smoothness determination for the image of FIG. 25, the upper left horizontal edge, the upper right horizontal edge, the lower left horizontal edge, and the lower right horizontal are extracted as “smooth edge” as a result of the edge smoothness determination for the image of FIG. It will be an edge.
This is detected as a continuous edge and a pair edge, and finally identified as a “character edge” (unless excluded by the correction means 293).

(Specific example 2 of edge smoothness determination)
Next, edge smoothness determination in the case of a halftone dot pattern will be specifically described below.
FIG. 30 shows an image after binarization of the halftone dot pattern.
FIG. 31 shows the result of the P1 pattern matching unit 241 and the result of the horizontal pixel number counting unit 245 for the image of FIG. In the upper left vertical edge, the horizontal pixel count results are all less than 7 and are arranged in descending order from top to bottom, so the extraction result is “1”, and the lower left vertical edge is the horizontal pixel count result = 7. Therefore, the extraction result is “0”.

FIG. 32 shows the result of the P2 pattern matching unit 242 and the result of the horizontal pixel count unit 246 for the image of FIG. 30. The upper right vertical edge has the horizontal pixel count result = 7, and the extraction result is “ “0” and the lower right vertical edge are all less than 7 in the horizontal pixel count result and are arranged in descending order from top to bottom, so the extraction result is “1”.
FIG. 33 shows the result of the P3 pattern matching unit 243 and the result of the vertical pixel number counting unit 247 for the image of FIG. 30, and the upper left horizontal edge shows that the vertical pixel count results are all less than 7, but from left to right. Since the pixels are not arranged in descending order toward the left, the extraction result is “0”, and the upper right horizontal edge is all less than 7 in the vertical direction pixel count result and is arranged in descending order from left to right. 1 ”.
FIG. 34 shows the result of the P4 pattern matching unit 244 and the result of the vertical pixel number counting unit 248 for the image of FIG. 30, and the lower left horizontal edge shows that the vertical pixel count results are all less than 7 and the left to right The result of extraction is “1” because the pixels are arranged in descending order toward the left, and the lower right horizontal edge has all the vertical pixel count count results of less than 7, but is not arranged in descending order from left to right. It becomes “0”.

From the above extraction results, as a result of the edge smoothness determination for the image shown in FIG. 30, the upper left vertical edge, the lower right vertical edge, the upper right horizontal edge, and the lower left horizontal edge are extracted as “smooth edge”. The
This is not a continuous edge, a pair edge, or an intersection edge, and the final result is not a character edge. It should be noted here that the upper left horizontal edge and the lower right horizontal edge are not extracted because they are judged as non-smooth edges due to local irregularities in the reference area. This prevents it from being identified.

In the present embodiment, the above-described edge smoothness determination means 203, 213, 223, continuous edge detection means 204, 214, 224, pair edge detection means 205, 215, 225, intersection edge detection means 206, 216, 226, In addition, basic identification means 507, 508, and 509 are constituted by the character and pattern identification means 207, 217, and 227.
FIG. 35 is a block diagram showing the configuration of the supplementary identification means 501, 502, and 503.
The supplementary identification means 501, 502, and 503 are composed of a short line / small circle detection means 601, a cross pattern detection means 602, and a T-shaped pattern detection means 603, and detect the pixel of interest as a short line, small circle, cross, or T-shaped pixel. In this case, it is identified as a character by OR 604 and “1” is output. In other cases, “0” is output.

Since the basic identification means 507, 508, and 509 detect continuous character edges of a predetermined number of pixels or more, short lines that exist in isolation cannot be identified as characters.
In addition, since the basic identification means 507, 508, and 509 detect only continuous vertical edges and horizontal edges of a predetermined number of pixels or more, there is a problem that character portions such as small circles with large curvatures cannot be identified.
Therefore, the short line / small circle detecting means 601 can detect a round shape having a large curvature. Thus, unlike the basic identification means 507, 508, and 509, the supplementary identification means 501, 502, and 503 perform the identification process even when the number of pixels of the character is less than the predetermined number of pixels necessary for the identification process. Can do.

FIG. 36 is a block diagram showing the configuration of the short line / small circle detecting means 601.
The vertical continuous white pixel extracting means 604 outputs “1” when the pattern matches the pattern shown in FIG.
The horizontal continuous white pixel extracting means 605 outputs “1” when the pattern matches the pattern shown in FIG.
The diagonally continuous white pixel extracting means 606 outputs “1” when it matches any of the patterns in FIG.
The diagonal direction continuous white pixel extracting means 607 outputs “1” when it matches any of the patterns in FIG.
In the vertical continuous white pixel extracting unit 604, the horizontal continuous white pixel extracting unit 605, the diagonal continuous white pixel extracting unit 606, and the diagonal continuous white pixel extracting unit 607, the center is the target pixel, and the output is the target. This is performed as an extraction result for the pixel, and “0” is output when the result does not match.
Next, in the region extraction unit 608 surrounded by white pixels, if all of the following conditions (1) to (4) are satisfied, the region extraction unit 608 outputs “1” as a region surrounded by white pixels. If it does not hold, “0” is output.
(1) One or more of a1, a2, a3 and b1, b2, b3 in FIG. 41 each have a pixel of output “1” from the vertically connected white pixel extracting means 604.
(2) One or more of c1, c2, c3 and d1, d2, d3 in FIG. 41 each have a pixel of output “1” from the laterally connected white pixel extracting means 605.
(3) One or more of e1, e2, e3 and f1, f2, f3 in FIG. 41 each have a pixel of output “1” from the diagonally connected white pixel extracting means 606.
(4) One or more of each of g1, g2, g3 and h1, h2, h3 in FIG. 41 has a pixel of output “1” from the diagonally connected white pixel extracting means 607.

FIG. 43 is a flowchart showing a processing procedure for counting the number of black pixels adjacent to the target pixel in the black pixel block counting means 609.
First, the black pixel block counting unit 609 determines whether or not the target pixel e [0] [0] is a black pixel (step 5).
When the target pixel e [0] [0] is a black pixel (step 5; Y), the pixel count at x = y = 0 is set to 1 (step 10).
Next, x + 1 adjacent pixels are recognized from the target pixel (step 15), and it is determined whether x is a pixel of 8 or less (step 20). When x is 8 or less (step 20; Y), e [x] [0] is set as an adjacent pixel, and it is determined whether or not the adjacent pixel e [x] [0] is a black pixel (step 25).
When the adjacent pixel e [x] [0] is a black pixel (step 25; Y), 1 is added to the count number (step 30).

Next, 1 is added to y (step 35), and adjacent pixels satisfying the condition of (x = odd number and y ≦ 3) or (x = even number and y ≦ 1) are extracted (step 40).
When adjacent pixels satisfying the condition are extracted (step 40; Y), it is determined whether or not the pixel extracted in step 40 is a black pixel (step 45). If it is a black pixel (step 45; Y), Assuming that the outer pixel e [x] [y] is a black pixel, 1 is added to the count number (step 50), and the process returns to step 35 described above.
If no adjacent pixel satisfying the condition is extracted in step 40, y = 0 is set (step 55), and the process returns to step 20. Further, if there is no pixel satisfying the condition in step 20 (step 20). N), the black pixel block counting means 609 ends the process of counting the number of black pixels.

If the result of the black pixel block count of the target pixel is equal to or greater than a predetermined threshold value (set to a value of about 8 to 12), the short line black pixel block determination unit 610 determines that it is a short line black pixel block. "1" is output, and if it is less than the predetermined threshold, "0" is output.
The black pixel block determining unit 612 for small circles outputs the output result from the black pixel block counting unit 609 when the target pixel E shown in FIG. 40 is 0 (that is, the target pixel is a white pixel), and the pixel A, If it is greater than or equal to a predetermined threshold value set in advance in B, C, and D, it is determined that it is a small black pixel block, and “1” is output.
The small circle black pixel block judging means 612 extracts a pixel in which white is present in the central pixel and a black block is present in a donut shape around it.
The predetermined threshold in the determination by the small circle black pixel block determining unit 612 may be set to a value that is about 2 to 4 smaller than the predetermined threshold in the short line black pixel block determining unit 610.

The AND 613 outputs “1” when the output result of the short line black pixel block determination unit 610 and the result of the extraction unit 608 of the region surrounded by the white pixel block are both “1”.
Thereby, a black block surrounded by white pixels can be extracted.
The delay adjusting unit 611 delays the result of the short line black pixel block determining unit 610 by three lines and inputs the result to the AND 613 in order to synchronize the position of the target line.
The expansion unit 615 refers to the 13 × 13 pixels around the pixel of interest, and outputs “1” when “1” is input to the AND 613 at least.

The AND 614 outputs “1” when both the output result of the black pixel block determining unit 612 for small circles and the result of the extracting unit 608 of the area surrounded by the white pixel block are “1”, otherwise Outputs “0”.
That is, it is possible to determine a pixel in which a white image is present at the center, a black lump is present around the white image, and the outside is surrounded by white.
The expansion means 616 refers to the surrounding 13 × 13 pixels of the target pixel, and outputs “1” when “1” is input to the AND 614 at least.

The black isolated point extracting unit 620 regards the target pixel as a black isolated point pixel and outputs “1” when all of the following conditions (1) to (8) are satisfied.
If even one does not hold, “0” is output.
(1) It conforms to the pattern (a-1) or (a-2) shown in FIG.
(2) It conforms to the pattern (b-1) or (b-2) shown in FIG.
(3) It conforms to the pattern (c-1) or (c-2) shown in FIG.
(4) It conforms to the pattern (d-1) or (d-2) shown in FIG.
(5) It conforms to the pattern (e-1) or (e-2) shown in FIG.
(6) It conforms to the pattern (f-1) or (f-2) shown in FIG.
(7) It conforms to the pattern (g-1) or (g-2) shown in FIG.
(8) It conforms to the pattern (h-1) or (h-2) shown in FIG.
In the pattern of FIG. 45, all the centers are set as the target pixel.

The white isolated point extracting unit 619 outputs “1” by regarding the target pixel as the white isolated point pixel when at least six of the following conditions (1) to (8) are satisfied, and even one of them is not satisfied. In this case, “0” is output.
Here, the case where at least six consecutive conditions are satisfied means that all but (1) and (2) below are satisfied, or all but (2) and (3) below are satisfied, or the following (3) All except for (4) are established, all of the following (4) and (5) are established, or all of the following (5) and (6) are established, or the following (6) (7 All other than (7) and (8) below are established, or all other than (1) and (8) below are established.
(1) It conforms to the pattern of (a-1) or (a-2) in FIG.
(2) It conforms to the pattern of (b-1) or (b-2) in FIG.
(3) It conforms to the pattern of (c-1) or (c-2) in FIG.
(4) It conforms to the pattern of (d-1) or (d-2) in FIG.
(5) It conforms to the pattern of (e-1) or (e-2) in FIG.
(6) It conforms to the pattern (f-1) or (f-2) in FIG.
(7) It conforms to the pattern (g-1) or (g-2) in FIG.
(8) It conforms to the pattern of (h-1) or (h-2) in FIG.
In the pattern of FIG. 46, all the centers are set as the target pixel.

The white isolated point extracting means 619 includes not only the white isolated point (1) in FIG. 47 but also white that includes a U-shaped shape that is not completely surrounded by black, such as the white isolated point (2). It is possible to determine an isolated point.
The checker pattern extraction unit 621 outputs “1” as the checker pattern pixel when it matches the pattern shown in FIG.
In either case, the center is the target pixel, and the output is the extraction result for the target pixel. If it does not match, “0” is output.
The short line picture area determination means 622 counts the number of black isolated points, white isolated points, and checker pattern pixels within 11 × 11 pixels, and a predetermined number (about 0 to 2) of pixels of black isolated points is preset. ) As described above, when there are one or more pixels of white isolated points or one or more checker pattern pixels, “1” is output as a pattern area, and “0” is output in other cases. .

The short line picture area determination means 622 is for canceling erroneous detection in the picture part in the short line detection, and utilizes the fact that there is no white isolated point or checker pattern shown in FIG.
Black isolated points do not exist as long as the binarization means 202, 212, and 222 can accurately convert the background area into white pixels, but the criteria may be relaxed during the conversion, and there may be remaining black isolated points. is there.
The expansion means 624 outputs “1” as long as there is at least one output “1” of the short line picture area determination means 622 within 13 × 13 pixels. If none exists, “0” is output.

The small circle picture area determination means 623 counts the number of black isolated points and checker pattern pixels within 11 × 11 pixels, and the number of black isolated point pixels is equal to or greater than a predetermined number (about 0 to 2), or When there is one or more checker pattern pixels, “1” is output as a picture area, and “0” is output in other cases.
Similar to the short line, this is for canceling erroneous detection in the pattern portion in small circle detection.
The expansion means 625 outputs “1” if there is at least one output “1” of the small circle picture area determination means 623 within 13 × 13 pixels. If none exists, “0” is output.
In the AND 617, when the output of the expansion means 615 is “1” and the output of the inverting means 626 is “1” (= the output of the expansion means 624 is “1”), the AND 617 is surrounded by white that exists in isolation. “1” is output as a short line, and “0” is output otherwise.
In the AND 618, when the output of the expansion means 616 is “1” and the output of the inverting means 627 is “1” (= the output of the expansion means 625 is “1”), the AND 618 is surrounded by white that exists in isolation. “1” is output as a small circle, and 0 is output otherwise.

FIG. 49 is a block diagram showing the configuration of the cross pattern detecting means 602. As shown in FIG. Since the basic identification means 507, 508, and 509 detect a continuous character edge of a predetermined number of pixels or more, a portion where a short vertical line and a horizontal line cross inside a crowded character cannot be identified as a character. . It is for detecting it.
When the vertical black pixel pattern matching unit 628 matches the pattern shown in FIG. 50, the vertical black pixel pattern matching unit 629 sets the pixel of interest output “1” as a vertical black pixel, and the horizontal black pixel pattern matching unit 629 If it matches the pattern shown in FIG. 4, the pixel of interest is output “1” as a black pixel in the horizontal direction, and “0” is output if it does not match.
If the vertical white pixel pattern matching unit 630 matches the pattern shown in FIG. 52, the vertical white pixel pattern matching unit 631 outputs “1” as the vertical white pixel, and the horizontal white pixel pattern matching unit 631 shows that shown in FIG. If it matches the pattern, “1” is output as white pixels in the horizontal direction.
In either case, the center is the target pixel, and the output is the extraction result for the target pixel. If it does not match, “0” is output.

The vertical line width counting means 632 counts the number of continuous black A pixels in the vertical direction including the target pixel.
It is possible to count how many pixels the black bar pattern shown in FIG. 50 continues vertically, and to detect the width of the vertical line. The horizontal line width counting means 633 can count the number of vertical black pixels including the target pixel in the vertical direction.
The width can be detected by counting how many pixels the black bar pattern shown in FIG.

FIG. 56 is a flowchart showing the processing procedure for counting the number of black pixels continuous in the vertical and horizontal directions in the vertical line width counting means 632 and the black pixel block counting means 609.
A processing procedure for counting the number of black pixels continuous in the vertical and horizontal directions in the black pixel block counting means 609 will be described with reference to the flowchart of FIG.
First, it is determined whether or not the target pixel d [0] shown in FIG. 54 is a black pixel (step 100).
If the target pixel is a black pixel (step 100; Y), the black pixel is counted in each of the two directions adjacent to the target pixel shown in FIG. 50 (step 105).
First, for the positive direction of m, the pixel d [m] necessary for counting is recognized, 1 is added to the current m number (step 110), and it is determined whether m is a pixel within 6 (step 115). .
Then, it is determined whether or not the pixel d [m] is a black pixel (step 120). When the pixel d [m] is a black pixel (step 120; Y), 1 is added to the count number (step 125).

On the other hand, for the negative direction of m, the pixel d [m] necessary for counting is recognized and 1 is subtracted from the current m number (step 130) to determine whether m is a pixel within -6 (step 130). 135). Then, it is determined whether or not the pixel d [m] is a black pixel (step 140). When the pixel d [m] is a black pixel (step 140; Y), 1 is added to the count number (step 145).
When m exceeds 6, or when m is less than −6 (step 115; N and step 135; N), the count values counted in step 125 and step 145 are summed, and 1 is added. A final count value is set (step 150). The final count value obtained is set as the Bk_t value (step 155).

The distance calculation means 634 and 635 between the white areas for crosses are inputted with white pixels in the vertical direction or the horizontal direction, respectively, and calculate the distance between the white A pixels shown in the following (1) to (4) shown in FIG. calculate.
(1) Distance L1 between the white A pixel existing in the region <1> and the white A pixel existing in the region <2>
(2) The distance L2 between the white A pixel existing in the region <3> and the white A pixel existing in the region <4>
(3) The distance L3 between the white A pixel existing in the region <1> and the white A pixel existing in the region <3>
(4) Distance L4 between the white A pixel existing in the region <2> and the white A pixel existing in the region <4>
However, the white A pixels for calculating the distance are limited to those existing on the same row or the same column. Also, if there is something close to the center row or column that matches the condition, priority is given.

FIG. 58 is a flowchart illustrating the processing of the distance calculation means 634 and 635 between the white areas for crosses. A processing procedure of the distance calculation means 634 and 635 between the white areas for crosses will be described with reference to the flowchart of FIG.
First, it moves to the row of h = 1 shown in FIG. 57 (step 200), sets m values of two points to m1 = 0 and m2 = 0 (step 205), and subtracts 1 from the m value of m1 (step 200). 210), whether m1 is −5 or more is determined (step 215). If the result of determination in step 215 is greater (step 215; Y), it is determined whether there is a black pixel in the column m1 (step 220).

Next, when there is a black pixel in step 220 (step 220; Y), 1 is added to the m value of m2 (step 225), and it is determined whether m2 is 5 or more (step 230). If the result of determination is below (step 230; Y), it is determined whether there is a black pixel in the column m2 (step 235). If there is a black pixel in step 235 (step 235; Y), the distance L1 is calculated by adding | m1 | and | m2 | (step 240).
In step 215 and step 230, if m1 is less than −5 and m2 is greater than 5 (step 215; N and step 230; N), the process moves to the line h + 1 (step 250), and h is 5 It is determined whether the following is true (step 255). If h is 5 or less, the process returns to step 205 (step 255; Y), and if h exceeds 5, the distance L1 is set to 0 (step 260).

The cross determination unit 636 and the cross determination unit 637 determine that the pixel is a cross pixel when all of the following conditions (1) to (3) are satisfied, and output “1”, otherwise "0" is output to.
(1) Vertical line width (Bk_t) ≠ 0, horizontal line width (Bk_y) ≠ 0, and four distances between white areas (L1, L2, L3, L4) are not all zero.
(2) The line widths of the vertical and horizontal lines are substantially constant: | L1-L2 | <th1 and | L3-L4 | <th1.
(3) A vertical line or a horizontal line exists between the white A pixel and the white A pixel: min (L1−L2) ≦ (Bk_t + th2) and min (L3−L4) ≦ (Bk_y + th2).
Here, th1 and th2 are predetermined threshold values set in advance, and are for giving a slight allowance to the criterion. It is set to a small value of about 2 to 4.

The cross determination means 636 can detect the cross pattern portion even when the white A pixel is vertically long (particularly 1 dot width) as shown in FIG.
The cross determination means 637 can detect the cross pattern portion even when the white A pixel is horizontally long (particularly, one dot width) as shown in FIG.
FIG. 60 is an enlarged view of the cross pattern portion at the center of FIG. The distances L1, L2, L3, and L4 between the white areas, and the line width Bk_t of the vertical line and the line width Bk_y of the horizontal line correspond to the distances of the illustrated portions in the cross pattern portion of FIG.
When the cross determining unit 636 or the cross determining unit 637 determines that the pixel is a cross portion pixel, the OR 638 outputs “1”, and the dilating unit 639 outputs “1” within 15 × 15 pixels. "Is present, the target pixel is finally regarded as a cross pattern pixel, and" 1 "is output.

FIG. 61 is a block diagram showing a configuration of the T-shaped pattern detection means 603. As shown in FIG. Since the basic identification means 507, 508, and 509 detect a continuous character edge of a predetermined number of pixels or more, the vertical line and the horizontal line existing inside the crowded character intersect with the T-shape, and either direction If the line is short, it cannot be identified as a character. Below, each structure of the T-shaped pattern detection means 603 is demonstrated in detail.
Vertical black pixel pattern matching means 640, horizontal black pixel pattern matching means 643, vertical white pixel pattern matching means 644, horizontal white pixel pattern matching means 646, vertical line width counting means 648, horizontal line width counting The means 649 has the same configuration as the cross pattern detection means 602 and will not be described.
The vertical black pixel pattern matching unit 642 sets the output of the pixel of interest to “1” when it matches the pattern shown in FIG. The horizontal black pixel pattern matching unit 641 sets the output of the pixel of interest to “1” as a black pixel when it matches the pattern shown in FIG. If it does not match, 0 is output.

The vertical white pixel pattern matching unit 645 sets the output of the target pixel to “1” when it matches the pattern shown in FIG. 64, and the horizontal white pixel pattern matching unit 647 matches the pattern shown in FIG. The output of the pixel of interest is “1”, assuming that it is a white B pixel. If it does not match, “0” is output.
The vertical line width counting unit 648 and the horizontal line width counting unit 651 have the same configuration as the vertical line width counting unit 632 and the horizontal line width counting unit 633 of the cross pattern detection unit 602, and therefore will be described. Is omitted.
However, the reference areas of the vertical line width count 648 and the horizontal line width count 651 are d [−3] to d [3] in FIG. 54 and FIG. In the line width count flow diagram of FIG. 56, the conditional branching expressions of 660 and 661 are changed to m ≧ −3 and m ≦ 3 and applied.

The distance calculating means 652 between the vertical T-shaped white areas receives the white A pixel and the white B pixel in the vertical direction, and calculates the following distances (1) to (4) shown in FIG.
(1) The distance L3 between the white A pixel existing in the region <1> and the white A pixel existing in the region <3>
(2) The distance L4 between the white A pixel existing in the region <2> and the white A pixel existing in the region <4>
(3) The distance L5 between the white B pixel existing in the region <5> and the column where L4 ≠ 0 is calculated
(4) The distance L6 between the white B pixel existing in the region <6> and the column for which L3 ≠ 0 is calculated
The distance calculation means 653 between the white areas for the T direction in the horizontal direction inputs the white A pixel and the white B pixel in the horizontal direction, and calculates the following distances (5) to (8) shown in FIG.
(5) The distance L1 between the white A pixel existing in the region <1> and the white A pixel existing in the region <2>
(6) The distance L2 between the white A pixel existing in the region <3> and the white A pixel existing in the region <4>
(7) Distance L7 between the white B pixel existing in region <7> and the row where L2 ≠ 0 is calculated
(8) Distance L8 between the white B pixel existing in region <8> and the row where L1 ≠ 0 is calculated

FIG. 68 is a flowchart showing the distance calculation processing of the distance calculating means 652 between the vertical T-shaped white areas and the distance calculating means 653 between the horizontal T-shaped white areas.
First, it is determined whether or not L1_h is 0 (step 300). When L1_h = 0 (step 300; Y), h = 0 is set (step 305), and 1 is subtracted from the current value. (Step 310), it is determined whether h is −5 or more (step 315).
If it is −5 or more (step 315; Y), W [h] is determined as 1 (step 320; Y), and L8 is set to (h + L1_h) (step 325).
If L1_h = 0 is not satisfied in step 325 (step 300; N), or if h is less than −5 in step 315 (step 315; N), L8 is set to 0 (step 330).

The calculation of L1, L2, L3, and L4 is the same as the distance calculation means 634 and 635 between the white areas for crosses. In the calculation of L5, L6, L7, and L8, for example, in L8, the line number h in which L1 ≠ 0 is calculated is calculated as L1_h. L5, L6, and L7 are the same as the other processes except that L1_h is changed to L4_h, L3_h, and L2_h. Whether or not the pixel is the white B pixel is always checked in order from the position closest to the target pixel, and the distance to the white B pixel closest to the target pixel in each reference region is calculated.
The T-shaped determination unit 654 determines that the following conditions (1-1), (2-1), and (3-1) are all satisfied, or (1-2), (2-2), and (3 -2) determines that the pixel is a T-shaped pixel and outputs "1", otherwise outputs "0".

(Determination condition of the T-shaped determination means 654)
(1-1) Vertical line width (Bk_t) ≠ 0, horizontal line width (Bk_y) ≠ 0, and two distances between white areas (L4, L5) are not zero.
(2-1) A horizontal line exists between the white A pixel and the white A pixel: L4 ≦ (Bk_y + th2)
(3-1) A vertical line exists between the white A pixel and the white B pixel: L5 ≦ (Bk_t + th2)
(1-2) Vertical line width (Bk_t) ≠ 0, horizontal line width (Bk_y) ≠ 0, and two distances between white areas (L3, L6) are not zero.
(2-2) A horizontal line exists between the white A pixel and the white A pixel: L3 ≦ (Bk_y + th2)
(3-2) A vertical line exists between the white A pixel and the white B pixel: L6 ≦ (Bk_t + th2)
th2 is a predetermined threshold value set in advance, and is the same as th2 used in the cross determination unit 636 and the cross determination unit 637.

  (1-1), (2-1), and (3-1) are T-shaped pattern portions shown in FIG. 69 (a), (1-2), (2-2), and (3-2). The part is a condition for detecting the T-shaped pattern part shown in FIG.

FIG. 70 is an enlarged view of the T-shaped pattern portion of FIG. 69 (a). The distances L4 and L5 between the white areas, and the line width Bk_t of the vertical line and the line width Bk_y of the horizontal line correspond to the distances of the illustrated portions in the T-shaped pattern portion of FIG.
The T-shaped determination unit 655 determines that the following conditions (1-1), (2-1), and (3-1) are all satisfied, or (1-2), (2-2), and ( When all of 3-2) are satisfied, it is determined that the pixel is a T-shaped pixel, and “1” is output. Otherwise, “0” is output.

(Determination condition of the T-shaped determination means 655)
(1-1) Vertical line width (Bk_t) ≠ 0, horizontal line width (Bk_y) ≠ 0, and two distances between white areas (L2, L7) are not zero.
(2-1) A vertical line exists between the white A pixel and the white A pixel: L2 ≦ (Bk_t + th2)
(3-1) A horizontal line exists between the white A pixel and the white B pixel: L7 ≦ (Bk_y + th2)
(1-2) The vertical line width (Bk_t) ≠ 0, the horizontal line width (Bk_y) ≠ 0, and the two distances between the white areas (L1, L6) are not zero.
(2-2) A vertical line exists between the white A pixel and the white A pixel: L3 ≦ (Bk_t + th2)
(3-2) A horizontal line exists between the white A pixel and the white B pixel: L6 ≦ (Bk_y + th2)
Note that th2 is a predetermined threshold value set in advance, and is the same as th2 used in the cross determination unit 636 and the cross determination unit 637.

When the T-character determining unit 654 or the T-character determining unit 655 determines that the pixel is a T-shaped portion, the OR 656 outputs “1”, and the dilating device 657 outputs even one in 15 × 15 pixels. When “1” exists, the target pixel is finally regarded as a T-shaped pattern pixel and “1” is output.
FIG. 71 is a block diagram showing the configuration of the edge amount extraction means 30. As shown in FIG. The signal synthesis means 301 synthesizes the R′G′B ′ signal ((R ′ + 2G ′ + B ′) / 4) and converts it into one signal, and the edge quantity detection means 302 detects the edge quantity.
FIG. 72 is a block diagram of the edge amount detection means 302, which is an edge amount detection filter, and performs masking calculation using the four types of 7 × 7 filters shown in FIG. Of the four outputs, the one with the maximum absolute value is selected by the maximum value selection 311 and output.

FIG. 74 is a block diagram showing the configuration of the filter processing means 13. The smoothing means 131 and the edge enhancement means 132 are subjected to filter processing, and the results of the two filter processes are combined by the combining means 133 at a ratio corresponding to the edge amount and the character edge. In the case of a character edge, the smoothing result: edge enhancement result is combined at a ratio of 1:10 when the edge amount is maximum according to the edge amount, and 1: 0 (only the smoothing result is valid) when the edge amount is minimum. When the edge amount is an intermediate value, it is preferable to synthesize at an intermediate ratio, and when it is not a character edge, it is always preferable to synthesize at a ratio of 1: 0.
In this manner, the white ground character edge and the halftone dot character edge are clearly reproduced with the edge emphasized, and the pattern portion is effective for reproducing smoothly and with good graininess.

FIG. 75 is a block diagram showing the configuration of the edge amount extraction means 40. As shown in FIG. C′M′Y ′ Multi-value edge amounts are detected by edge amount detection means 401, 402, and 403 for each color. The configuration of the edge amount detection means is the same as that of the edge amount extraction means 30 (FIG. 72).
As the edge amount detection filter, a filter of 5 × 5 size shown in FIG. 76 is used.
Then, the edge amount from the signal after filtering is detected, and the size of the detection filter is changed in consideration of the difference in resolution.
The maximum value selection unit 404 selects the maximum of the detected three color edge amounts. On the other hand, the signal synthesizing unit 405 adds the three signals C′M′Y ′ at a predetermined ratio and converts them into one signal. For example, (C ′ × 1/4 + M ′ × 2/4 + Y ′ × 1/4) is output as a synthesized signal. The Laplacian filter calculation means 406 performs a masking calculation on the synthesized signal using the filter shown in FIG. 77, and the sign determination means 407 determines whether the value is a positive value. At this time, “1” is output if positive, and “0” is output if negative.

When the result of the sign determination unit 407 is “1” by the inner edge amount extraction unit 408, the output value of the maximum value selection unit 404 is output as it is. If the result of the code determination 407 is “0”, the edge amount is canceled and “0” is output.
The output value of the Placian filter calculation means 406 outputs a positive value on the high density side of the edge corresponding to the inner edge of the character, and a negative value on the low density side of the edge corresponding to the outer edge of the character. , The edge amount is enabled only for the inner edge where the Laplacian is positive, and the edge amount is disabled for the outer edge where the Laplacian is negative.
An N value is quantized by the N value converting means 409. N is determined according to the number of stages of processing controlled by the UCR / black generation unit 15 to which the edge amount extraction result is applied, but here it is controlled in two stages to simplify the explanation. The inner edge amount extracted with a value of 0 to 63 is quantized to 0/1 (N = 2).

The UCR / black generation means 15 includes a method for performing both black generation and UCR using an equation, a method using an LUT, and the like. In this embodiment, a case where black generation is performed using an LUT and UCR is performed using an equation will be described.
FIG. 78 is a diagram showing a black generation LUT. In black generation, the minimum value Min (C ′, M ′, Y ′) of C ′, M ′, Y ′ (0 is white, 255 is black) is calculated and input to the LUT to obtain an output value. . UCR is performed by the following equation. α is a UCR adjustment parameter.
C = C′−α × K
M = C′−α × K
Y = C′−α × K (Formula 4)

  When the output value of the character edge and edge amount extraction means 40 is “1”, the black generation table-1 is used, Min (C, M′Y ′) is replaced with K as it is, and the UCR adjustment parameter is α = 1. To do. In other cases, the black generation table-2 is used, and K is not generated in the highlight, but K is gradually generated from the middle. The UCR adjustment parameter is α = 0.5. As in the filter processing, the edge amount may be obtained in multiple values, and a number of intermediate black generation tables may be prepared and controlled in multiple stages. By switching between black ink generation and UCR in this way, it is possible to reproduce characters (black characters) clearly and without color and to reproduce a pattern with high gradation.

As described above, according to the first embodiment, basic identification means for identifying a character edge based on smooth edges that are continuously present in excess of a predetermined number of pixels, a short line, a small circle, a cross pattern, and a T-character. In order to identify the character on the halftone dot based on both identification results by the supplementary identification means that identifies the pattern, the character on the halftone dot is separated from the edge of the halftone dot pattern and with high accuracy without missing in the character part Can be extracted.
Furthermore, by applying the identification result to filter processing and black ink processing, it is possible to achieve both high-quality reproduction of halftone dot character edges and halftone dot patterns. In the present embodiment, an example is shown in which the character / pattern identification result is applied to the filter processing and the black processing. However, a conventionally known method in which the character / pattern identification result is used for various image quality enhancement processing such as color correction and pseudo gradation processing. Of course, it is also possible to apply the identification result of the present invention to these techniques.

[Details of Second Embodiment]
In the first embodiment, an example in which the character / pattern identification result of the present invention is applied to the image quality improvement processing has been described. However, in the second embodiment, toner save processing that is not image quality improvement but image quality processing is performed. An example applied to The toner saving process is a process for suppressing the toner consumption amount by lowering the density and outputting. The “toner save” toner is not limited to powder toner. This includes liquid toner (ink).

FIG. 79 is a block diagram showing a configuration of the present embodiment. On the operation panel 50, when the user wants to output in the toner save mode, the toner save mode can be selected and designated. When the user designates the toner save mode, the toner save processing unit 19 performs toner save processing.
The toner save processing unit 19 performs toner save processing according to the signal from the operation panel 50, the result of the image attribute determination unit 20 (whether it is a character edge), and the output value of the edge amount extraction unit 40. Switch.
FIG. 80 is a diagram showing γ conversion in the toner save process. When toner save γ−1 is used, the input value = the output value, the density is stored as it is, and the toner save process is not substantially performed. When toner save γ-2 is used, processing is performed in which the output value is about 50% of the input value. Therefore, when the toner save mode is not designated on the operation panel, toner save γ-1 is always used. When the toner save mode is designated on the operation panel, the toner save process is performed with toner save γ−1 when the character edge is detected and the output value of edge amount extraction is “1” (substantially). Not to do). In other cases, toner save γ-2 is used.

  As a result, when the toner save mode is set, the density of the white background character edge and the character edge on the halftone dot is preserved to maintain the character legibility, while effectively saving the toner on the picture portion, and Generates a good toner-save image even in the picture part, suppressing image quality defects (faults that frequently occur when the accuracy of character / pattern identification is poor) where the edges of the dot pattern are conserved in density like character edges be able to.

  As described above, according to the second embodiment, by applying the character / picture identification means of the present invention to the toner saving process, it is possible to simultaneously maintain character legibility, suppress defects in the picture image quality, and exhibit a toner saving effect. Can be achieved. Thus, the identification technique of the present invention is not limited to image processing for improving image quality, and the application destination is extremely wide. The toner save mentioned here is one example of application, and in addition to this, for example, it can be effectively applied to a technique for increasing compression efficiency by identifying a character pattern and changing the compression method.

[Details of Third Embodiment]
FIG. 81 is a block diagram showing a configuration for carrying out the third embodiment. The third embodiment is characterized in that the basic identification means 514, 524, and 534 are different from the first embodiment in that basic identification is performed using another character feature.
The basic identification means 514, 524, 534 in the third embodiment will be described. In this embodiment, a ridge pixel is used as a character feature, and the ridge pixel sandwiches a pixel of interest as shown in FIG. 84 (a) or a feature that is continuously present inside the character, as shown in FIG. 84 (b). However, the pixel density on both sides is such a characteristic that the pixel is lower than the density of the target pixel by a predetermined value or more.
The smoothing means 511, 521, and 531 perform smoothing processing as preprocessing for improving the accuracy of ridge detection. For example, the filter shown in FIG. 82 is used, and the filter parameter is adjusted according to the MTF characteristic of the scanner.

The ridge detection means 512, 522, and 532 detect the target pixel Lc as a ridge pixel when three of the following conditions (1) to (4) are satisfied for the 5 × 5 pixels shown in FIG. 1 "is output. In other cases, “0” is output.
(1) Lc-L1 ≧ Th and Lc-L24 ≧ Th
(2) Lc-L3 ≧ Th and Lc-L22 ≧ Th
(3) Lc-L5 ≧ Th and Lc-L20 ≧ Th
(4) Lc-L11 ≧ Th and Lc-L14 ≧ Th

The density correction means 513, 523, and 533 refer to 21 × 21 pixels when the ridge pixels are continuously present, and therefore the predetermined threshold value that the ridge pixels are set in advance in the reference region. If it exists, it is identified as a character and “1” is output. If it is less than the predetermined threshold, “0” is output. It is appropriate to set the predetermined threshold to a value of 15 or more with respect to the reference area size of 21 × 21.
As described above, according to the third embodiment, the basic identification for identifying characters based on the ridge pixels continuously present with a predetermined number of pixels or more, and the short line, small circle, cross pattern, and T-shaped pattern are identified. Since the characters on the halftone dot are identified based on both identification results, the character on the halftone dot is separated from the middle edge of the halftone dot pattern, and the character portion can be extracted with high accuracy without omission. .

It is the block diagram which showed the structure of the 1st Embodiment of this invention. It is the block diagram which showed the structure of the image attribute determination means. It is the block diagram which showed the structure of the binarization means. It is the figure which showed 3 * 3 pixel and 5 * 5 pixel. It is the block diagram which showed the structure of the smoothness determination means of an edge. It is the figure shown about the pattern matching performed with respect to 3 * 3 size binary image data. It is the figure which showed the example which made extraction object the upper left vertical edge and the lower left vertical edge. It is the figure which showed the example which made extraction object the upper right vertical edge and the lower right vertical edge. It is the figure which showed the example which made extraction object the upper left horizontal edge and the upper right horizontal edge. It is the figure which showed the example which made extraction object the upper left horizontal edge and the lower right horizontal edge. It is the figure which showed the example which made extraction object the lower left vertical edge and the lower right vertical edge. It is a figure explaining the process of the horizontal direction pixel number count means. It is a figure explaining the process of the vertical direction pixel number count means. It is the figure which showed the reference area and extraction condition of an upper left vertical edge extraction means and an upper right vertical edge extraction means. It is the figure which showed the reference area and extraction condition of a lower left vertical edge extraction means and a lower right vertical edge extraction means. It is the figure which showed the reference area and extraction condition of an upper left horizontal edge extraction means and a lower left horizontal edge extraction means. It is the figure which showed the reference area and extraction condition of an upper right horizontal edge extraction means and a lower right horizontal edge extraction means. It is the block diagram which showed the structure of the continuous edge detection means. It is the block diagram which showed the structure of the pair edge detection means. It is the block diagram which showed the structure of the intersection edge detection means. It is the figure shown about the example of extraction of a continuous edge. It is the figure shown about the example of extraction of a pair edge. It is the figure shown about the example of extraction of an intersection edge. It is the block diagram which showed the structure of the character picture identification means. It is the figure which showed the image after binarization of the character edge on a halftone dot. It is the figure which showed the result of the P1 pattern matching means with respect to the image of FIG. 25, and the result of the horizontal pixel number counting means. It is the figure which showed the result of the P2 pattern matching means with respect to the image of FIG. 25, and the result of the horizontal pixel number counting means. It is the figure which showed the result of the P3 pattern matching means with respect to the image of FIG. 25, and the result of the vertical direction pixel number count means. It is the figure which showed the result of the P4 pattern matching means with respect to the image of FIG. 25, and the result of the vertical direction pixel number count means. It is the figure which showed the image after the binarization of the halftone dot pattern. It is the figure which showed the result of the P1 pattern matching means with respect to the image of FIG. 30, and the result of the horizontal pixel number counting means. It is the figure which showed the result of the P2 pattern matching means with respect to the image of FIG. 30, and the result of the horizontal pixel number counting means. It is the figure which showed the result of the P3 pattern matching means with respect to the image of FIG. 30, and the result of a vertical direction pixel number count means. It is the figure which showed the result of the P4 pattern matching means with respect to the image of FIG. 30, and the result of the vertical direction pixel number count means. It is the block diagram which showed the structure of the supplement identification means. It is the block diagram which showed the structure of the short line / small circle detection means. It is the figure shown about the pattern at the time of a vertical direction continuous white pixel extraction means extracting. It is the figure shown about the pattern at the time of a horizontal direction continuous white pixel extraction means extracting. It is the figure shown about the pattern at the time of a diagonal direction continuous white pixel extraction means extracting. It is the figure shown about the pattern at the time of a diagonal direction continuous white pixel extraction means extracting. It is the figure which showed the conditions of the area | region enclosed by the white pixel in the extraction means of the area | region enclosed by the white pixel. It is the figure shown about each pixel arrange | positioned at 5x5. It is the flowchart which showed the process sequence which counts the pixel number of the black pixel adjacent to the attention pixel in a black pixel block count means. It is the figure shown about the criteria determined about a black pixel of a small circle. It is the figure shown about the reference | standard which determines the black isolated point of a black isolated point extraction means. It is the figure shown about the reference | standard which determines the white isolated point of a white isolated point extraction means. It is the figure shown about the white isolated point which a white isolated point extraction means extracts. It is the figure shown about the pattern at the time of a checker pattern extraction means extracting. It is the block diagram which showed the structure of the cross pattern detection means. It is the figure shown about the pattern at the time of a vertical direction black pixel pattern matching means adapting. It is the figure shown about the pattern at the time of a horizontal direction black pixel pattern matching means adapting. It is the figure shown about the pattern at the time of a vertical direction white pixel pattern matching means adapting. It is the figure shown about the pattern at the time of a horizontal direction white pixel pattern matching means adapting. It is the figure which showed the pixel of the horizontal direction. It is the figure which showed the pixel of the vertical direction. It is the flowchart which showed the process sequence which counts the pixel number of the black pixel which continued in the vertical direction and horizontal direction in the line width count means of a vertical line, and a black pixel block count means. It is the figure shown about the white A pixel of each area | region. It is the flowchart which showed the distance calculation procedure of the distance calculation means between the white areas for crosses. It is the figure which showed the vertical length and the horizontal length of the white A pixel at the time of detecting a cross pattern. It is the figure which expanded the cross pattern of Fig.59 (a). It is the block diagram which showed the structure of the T-shaped pattern detection means. It is the figure shown about the pattern at the time of a vertical direction black pixel pattern matching means adapting. It is the figure shown about the pattern at the time of a horizontal direction black pixel pattern matching means adapting. It is the figure shown about the pattern at the time of a vertical direction white pixel pattern matching means adapting. It is the figure shown about the pattern at the time of a horizontal direction white pixel pattern matching means adapting. It is the figure shown between the white pixels of the vertical direction which the distance calculation means between the white areas for vertical T characters calculates a distance. It is the figure shown between the white pixels of the horizontal direction which the distance calculation means between the white areas for horizontal T characters calculates a distance. It is the flowchart which showed the distance calculation procedure of the distance calculation means between the white areas for vertical T characters, and the distance calculation means between the white areas for horizontal T characters. It is the figure which showed the example of the T-shaped pattern. It is the figure which expanded the T-shaped pattern of Fig.69 (a). It is the block diagram which showed the structure of the edge amount extraction means. It is the block diagram which showed the structure of the edge amount detection means. It is the figure which showed four types of 7 * 7 filters used by a masking calculation. It is the block diagram which showed the structure of the filter process means. It is the block diagram which showed the structure of the edge amount extraction means. It is the figure which showed the edge amount detection filter. It is the figure which showed an example of the filter used when a Laplacian filter calculating means calculates with respect to a synthesized signal. FIG. 6 is a diagram showing a black generation LUT. It is the block diagram which showed the structure of the 2nd Embodiment of this invention. FIG. 6 is a diagram illustrating γ conversion in a toner save processing unit. It is the block diagram which showed the structure of the 3rd Embodiment of this invention. It is the figure shown about the filter used for the pretreatment of a ridge detection. It is the figure shown about the pattern at the time of a ridge detection means detecting a ridge. It is the figure shown about an example of the ridge pixel. It is the figure shown about the halftone dot pattern edge binarized with threshold value Th1, and the character edge on a halftone dot. It is the figure shown about the halftone dot pattern edge binarized with threshold value Th2, and the character edge on a halftone dot. It is the figure which showed the cloud point, the semi-cloudy point, the cross part, and the T-shaped part.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Scanner 11 Scanner gamma correction means 12 Scanner color correction means 13 Filter processing means 14 Printer color correction means 15 UCR / black generation means 16 Printer gamma correction means 17 Pseudo gradation processing means 18 Printer 20 Image attribute judgment means 30, 40 Edge amount Extraction means

Claims (10)

Image data receiving means for receiving image data;
A first pattern matching process, a horizontal pixel count process, and a vertical pixel count process are performed on the image data received by the image data receiving unit , and the number of pixels in which black pixels or white pixels are continuous is determined. Basic identification means for identifying character characteristics of a character on a halftone dot having a predetermined number of pixels or more;
For the image data received by the image data receiving means, a continuous white pixel extraction process, an extraction process of a region surrounded by white pixels, a black pixel block count process, an isolated point extraction process, a checker pattern extraction process, a second The pattern matching process and the third pattern matching process are performed, and the number of consecutive black pixels or white pixels different from the character characteristics of the halftone dot character identified by the basic identification means is less than the predetermined pixel number. Supplementary identifying means for identifying character features of non-halftone characters as character features of halftone characters;
Based on the character characteristics of the halftone dot character identified by the basic identification means and the character characteristic of the halftone dot character identified by the supplementary identification means, the image data received by the image data receiving means is defined as a character area. An image processing apparatus comprising: an image attribute determining unit that separates a non-character area.   The image processing apparatus according to claim 1, wherein the basic identification unit identifies a character feature of the character on the halftone dot as an edge portion.   The basic identification means identifies a character feature of a character on a halftone dot in which the number of pixels in which black pixels or white pixels are continuous is equal to or greater than a predetermined number of pixels in a certain direction. The image processing apparatus described.   The image processing apparatus according to claim 1, wherein the basic identification unit identifies a character feature of the character on the halftone dot as a ridge pixel.   The image processing apparatus according to claim 1, wherein the supplementary identification unit identifies a character feature of the character on the halftone dot as an isolated part.   The supplementary identifying means identifies the character feature of the character on the halftone dot as a part forming a round shape, 6. The claim 1, claim 3, claim 4, or claim 5, wherein The image processing apparatus described.   The supplementary identifying means identifies the character feature of the character on the halftone dot as a portion where a line crosses a cross, 5. Claims 1, 2, 3, 4, 5. Alternatively, the image processing apparatus according to claim 6.   The supplementary identifying means identifies a character feature of the character on the halftone dot as a portion where a line intersects a T-character. The image processing apparatus according to claim 6 or 7. In an image processing apparatus provided with image data receiving means for receiving image data,
A first pattern matching process, a horizontal pixel count process, and a vertical pixel count process are performed on the image data received by the image data receiving unit , and the number of pixels in which black pixels or white pixels are continuous is determined. A first step of identifying a character feature of a character on a halftone dot having a predetermined number of pixels or more;
For the image data received by the image data receiving means, a continuous white pixel extraction process, an extraction process of a region surrounded by white pixels, a black pixel block count process, an isolated point extraction process, a checker pattern extraction process, a second The pattern matching process and the third pattern matching process are performed, and the number of continuous black pixels or white pixels different from the character characteristics of the halftone character identified in the first step is set to the predetermined number of pixels. A second step of identifying character features of less than halftone characters as character features of halftone characters;
Based on the character feature of the halftone dot character identified in the first step and the character feature of the halftone dot character identified in the second step, the image data received by the image data receiving means is a character. And a third step of separating the region into a non-character region. An image data receiving function for receiving image data;
The first pattern matching process, the horizontal pixel count process, and the vertical pixel count process are performed on the image data received by the image data reception function , and the number of pixels in which black pixels or white pixels are continuous is determined. A basic identification function for identifying character features of halftone characters that are equal to or greater than a predetermined number of pixels;
For the image data received by the image data receiving function , continuous white pixel extraction processing, extraction processing of a region surrounded by white pixels, black pixel block counting processing, isolated point extraction processing, checker pattern extraction processing, second The pattern matching process and the third pattern matching process are performed, and the number of consecutive black pixels or white pixels, which is different from the character characteristics of the character on the halftone dot identified by the basic identification function, satisfies the predetermined number of pixels. A supplementary identification function that identifies character features of non-halftone characters as character features of halftone characters;
Based on the character feature of the halftone dot character identified by the basic identification function and the character feature of the halftone dot character identified by the supplementary identification function, the image data received by the image data receiving function is defined as a character region. A computer-readable storage medium storing an image processing program, comprising: an image attribute determination function that separates into a non-character area.

Images