JP2006262215A - Image processor, image processing method, image processing program and recording medium with the program recorded thereon - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve black character quality in particular and to improve image quality by lowering the probability of erroneously determining a black character rather than a color character. <P>SOLUTION: A dot area detecting part 263 receives gpk, bpk and iro. At iro=1, that is, when a pixel under consideration is determined to be chromatic, an OR operation of gpk and bpk is regarded as a dot peak pixel, the number of dot peak pixels is counted in each small two-dimensional area of a prescribed size, and the total sum of dot peak pixels of a mountain and a valley is defined as a count value P. At iro=0, that is, when a pixel under consideration is determined to be achromatic, an AND operation of gpk and bpk is regarded as a dot peak pixel, the number of dot peak pixels is counted in each small two-dimensional area of a prescribed size, and the total sum of the dot peak pixels of a mountain and a valley is defined as a count value P. When the count value P is larger than a threshold Pth, the small two-dimensional area is determined to be a dot area. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image processing apparatus and an image processing method having a function of receiving input of image data, performing predetermined image processing on the input image data, and outputting the input image data, such as a copying machine and an MFP (multi-function printer) The present invention relates to an image processing program and a recording medium on which the program is recorded.

2. Description of the Related Art Conventionally, there is a color image processing apparatus that includes an image area separation unit that determines whether each image area is a character area or a picture area for an image read from an image reading unit, and performs different image processing.
In such an image processing apparatus, when black characters and color characters are considered, if the black characters are colored due to the RGB reading position shift of the scanner or the plate misregistration occurs in each color of the plotter, the legibility is significantly deteriorated. Therefore, for black characters, a process of outputting Bk single color data based on a signal corresponding to luminance is performed, and this process is called black character processing.

  In addition, as a conventional technique for performing such image area separation, the technique disclosed in Patent Document 1 previously filed by the applicant of the present invention is based on the feature information in the vicinity of the local area of interest of the dot area detecting means. The threshold used for detecting the point area is made variable.

Further, in the case of Patent Document 2 previously filed by the present applicant, the image recognition means includes a white background detection means for detecting a low density area of an image and a black background detection means for detecting a high density area. Accordingly, there is provided means for applying a process for improving the image quality to the image data.
JP-A-5-29212 JP 2001-292316 A

  In the above-described conventional image processing apparatus, the number of distributions in a small area of halftone dot peak pixels is counted for the original image data read by the scanner, and the counted value is compared with a threshold value. It was determined whether or not. Although a threshold value was selected so as not to misdetermine the character at the time of this determination, in reality, a small character of grade 7 or a reduced scaled character has characteristics common to the halftone dot region, There was a risk of erroneous determination.

  In the prior art, a signal sensitive to the black density of a color image among RGB input images is a G signal. Therefore, a G signal is often used as an input signal to the halftone dot region detection means. However, since the G signal is not sensitive to the yellow color of the document, the B signal is also used to determine the halftone dot. As a result, the detection accuracy of the halftone dot region is increased. However, in the conventional technique, if the halftone dot peak is detected in the G signal or the B signal at the time of the halftone dot coefficient, the target pixel is set as the halftone dot peak pixel. Therefore, there is a possibility that the number of halftone dots in the character area may increase due to the influence of the RGB reading position shift of the scanner.

  When the black character is erroneously determined in this way, the present inventor does not perform the above-described black character processing, and the character quality is remarkably deteriorated. Therefore, in order to improve the image quality by image area separation, the present inventor makes an erroneous determination rather than the color character. It was found that the probability needs to be lowered.

  The above-mentioned Patent Documents 1 and 2 have not been taken into consideration to reduce the probability of erroneously determining a black character rather than such a color character.

  Therefore, the present invention provides an image processing device, an image processing method, an image processing program, and the program capable of improving the quality of black characters and improving the image quality by reducing the probability of erroneously determining a black character rather than a color character. An object of the present invention is to provide a recording medium on which is recorded.

  In order to achieve such an object, an image processing apparatus according to a first aspect of the present invention includes an image reading unit that reads an image, and a chromatic region that includes colors other than black and white from image data read by the reading unit. The chromatic area detecting means for detecting, and the two-dimensional small area in the image data read by the reading means, and the density difference between the pixel of interest in the two-dimensional small area and the surrounding pixel group around the pixel of interest Comparison determination means for determining whether or not the absolute value of each pixel is larger than a predetermined threshold value, and the target pixel determined by the comparison determination means that the absolute value of the density difference is larger than the threshold value. Halftone peak pixel detection means for forming halftone dot pixels forming part of a point, halftone dot peak pixels detected by the halftone dot pixel detection means, and halftone dot peak pixels detected around the halftone peak pixel A halftone dot area detecting means for detecting a two-dimensional small area of a predetermined size including the halftone dot peak pixel as a halftone dot area from the relationship, and the halftone dot area detecting means is detected by the chromatic area detecting means. The halftone dot detection result is changed according to the result.

  Among the color components included in the color image data read by the reading means described above, the first image data sensitive to the black density, and the second image having high sensitivity with respect to the color component having low sensitivity in the first image data. It is preferable that the image processing apparatus further includes color-use halftone area detection means for detecting a halftone area from color image data read by the reading means using at least two pieces of image data.

  The first image data described above is preferably G image data for each color component of RGB.

  The second image data described above is preferably B image data in each of RGB color components.

  The above-described halftone dot area detection means detects halftone dot peak pixel detection results in the first image data and halftone dot peak pixel detection in the second image data for the area determined as the chromatic area by the chromatic area detection means. It is preferable to detect a two-dimensional small region having a predetermined size as a halftone dot region using a logical sum with the result.

  The above-described halftone dot area detection means detects the halftone dot peak pixel detection result in the first image data and the halftone dot peak pixel detection in the second image data for the area determined as the achromatic area by the chromatic area detection means. It is preferable to detect a two-dimensional small region having a predetermined size as a halftone dot region using a logical product with the result.

  The image processing method according to the second aspect of the present invention detects an chromatic area including a color other than black and white from the image input process that receives the input of the image signal and the image data input by the image input process. Referring to the two-dimensional small region in the image data input by the chromatic region detection step and the image input step, the density difference between the pixel of interest in the two-dimensional small region and the surrounding pixel group around the pixel of interest A comparison / determination step for determining whether the absolute value is larger than a predetermined threshold value, and a pixel of interest determined by the comparison / determination step when the absolute value of the density difference is larger than the threshold value. A halftone dot pixel detection step that forms a halftone dot pixel that forms a part of the halftone dot pixel, a halftone dot peak pixel detected by the halftone dot peak pixel detection step, and a halftone dot peak pixel detected around the halftone dot peak pixel. From relationship A halftone dot region detecting step for detecting a two-dimensional small region of a predetermined size including the halftone dot peak pixel as a halftone dot region, and the halftone dot region detecting step corresponds to a detection result by the chromatic region detecting step. And changing the detection result of the halftone dot region.

  Among the color components included in the color image data input by the image input process described above, the first image data that is sensitive to the black density and the first image data that has a high sensitivity to the color components that are low in sensitivity. It is preferable to include a color-use halftone area detection step of detecting a halftone area from the color image data input by the image input step using at least two pieces of image data of the two image data.

  In the above-described halftone dot area detection step, the halftone dot peak pixel detection result in the first image data and the halftone dot peak pixel detection in the second image data for the area determined as the chromatic area by the chromatic area detection step. It is preferable to detect a two-dimensional small region having a predetermined size as a halftone dot region using a logical sum with the result.

  In the above-described halftone dot region detection step, the halftone dot peak pixel detection result in the first image data and the halftone dot peak pixel detection in the second image data for the region determined as the achromatic region by the chromatic region detection step. It is preferable to detect a two-dimensional small region having a predetermined size as a halftone dot region using a logical product with the result.

  An image processing program according to a third aspect of the present invention causes a computer to execute processing according to any of the steps described in any of the above-described image processing methods according to the second aspect of the present invention. .

  A recording medium recording an image processing program according to the fourth aspect of the present invention is characterized in that the image processing program according to the third aspect of the present invention described above is recorded.

  As described above, according to the present invention, the probability of erroneously determining a black character rather than a color character can be reduced. For this reason, especially the quality of black characters can be improved and the image quality can be improved.

  Next, an image processing apparatus, an image processing method, an image processing program, and a recording medium storing the program according to an embodiment of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing a schematic configuration of a digital full-color image processing apparatus according to the present embodiment. When operating as a full-color copying machine, the scanner 11 reads image data color-separated into R (red), G (green), and B (blue) from a document, and converts the image data (analog signal) into digital data. Output.

  As will be described later, the scanner correction unit 12 performs scanner γ correction processing on the RGB image data (digital data) read by the scanner 11, and classifies the image area into characters, line drawings, designs, etc. (image area separation). The characteristics of the scanner 11 are corrected by performing image processing such as emphasizing the character portion of the image and smoothing the pattern portion.

  The compression processing unit 13 compresses the multivalued image data after scanner correction, and sends the data to the general-purpose bus. The compressed image data is sent to the controller 14 through the general-purpose bus. The controller 14 has a semiconductor memory (not shown) and accumulates sent data.

The accumulated data is written to a hard disk drive (HDD) 15 which is a large capacity storage device as needed. This is because paper is jammed at the time of printout, so that it is possible to avoid re-reading the original even if the output does not end normally, to perform electronic sorting to rearrange multiple original image data, and to store the read original This is to re-output when necessary.
In the above description, the image data is compressed. However, if the general-purpose bus has a sufficiently wide bandwidth and the capacity of the HDD to be stored is large, the data may be handled in an uncompressed state.

  Next, the controller 14 sends the image data of the HDD 15 to the decompression processing unit 17 via the general-purpose bus. The decompression processing unit 17 decompresses the image data that has been subjected to compression processing to the original multi-value data, and sends it to the printer correction unit 18. The printer correction unit 18 performs printer gamma correction processing and gradation processing, and performs image correction processing, dithering processing, etc. according to the light-darkness correction processing of the plotter 19 and the gradation characteristics of the plotter 19 and the image area separation result. Data quantization is performed.

  The plotter 19 is a transfer paper printing unit using a laser beam writing process. The plotter 19 draws image data as a latent image on a photoconductor, and forms a copy image on the transfer paper after image formation / transfer processing with toner.

When operating as a distribution scanner that distributes image data to a PC via a network, an original image is read from the scanner 11 as in the case of operating as a copier, and from the scanner correction unit 12 through the compression processing unit 13, An image is sent to the controller 14. The controller 14 performs format processing. In the format processing, general-purpose image format conversion to JPEG, TIFF, or BMP format is performed.
Thereafter, the image data is distributed to an external PC terminal via a NIC (network interface controller) 16.

  The configuration of the scanner correction unit 12 will be described below. As shown in FIG. 2, the scanner correction unit 12 determines, based on the RGB image data input from the scanner 11, whether the image area (image area) of the document is a character area, a halftone character area, a design area, or the like. The image area separation unit 125, the scanner γ correction unit 121 that converts the digital value of the RGB image data according to the characteristics of the scanner 11 into a digital value proportional to the brightness, and the image data is sharpened based on the result of the image area separation A filter processing unit 122 that performs processing or smoothing processing, and image data that has been color-separated into RGB are converted into C (cyan), M (magenta), Y (yellow), A color correction unit 123 that converts to color image data including K (black) recording color information, and a scaling unit 124 that outputs the input image by enlarging or reducing the size in the main scanning direction. Configured to include a.

  The configuration of the printer correction unit 18 will be described below. As shown in FIG. 3, the printer correction unit 18 includes a printer γ correction unit 181 that performs γ correction on the CMYK image data that has passed through the expansion processing unit 17 according to the frequency characteristics of the plotter 19, and a dither processing / error. A gradation processing unit 182 that performs quantization such as diffusion processing, and outputs image data of each color of C ″ M ″ Y ″ K ″.

  An overall block diagram of the image area separation unit 125 is shown in FIG. The image area separation unit 125 is roughly divided into a filter unit 251, a white region extraction unit 253, a halftone dot separation unit 255, an edge extraction unit 252, a color determination unit 254, and an overall determination unit 256, and a 2-bit signal C / P and A 1-bit signal B / C is generated. These C / P and B / C signals are input to each processing unit of the scanner correction unit 12 in synchronization with RGB or CMYK color-separated image data, and are referred to as necessary. After the compression / decompression process is performed together with the image data, it is also input to the printer correction unit 18 and is referred to as necessary by each processing unit of the printer correction unit 18.

Next, the filter unit 251, the white region extraction unit 253, the edge extraction unit 252, and the color determination unit 254 of the image area separation unit 125 will be described using a configuration example provided with the document recognition unit 320 of FIG. 5.
That is, the filter unit 321, the white region extraction unit 323, the edge extraction unit 322, and the color determination unit 325 in the configuration example including the document recognition unit 320 of FIG. 5 are the filter units of the image area separation unit 125 in the embodiment of the present invention. 251, the white area extraction unit 253, the edge extraction unit 252, and the color determination unit 254 are similar to each other.
In the following description of the configuration example including the document recognition unit 320 of FIG. 5, a case where the reading density of the scanner is 600 dpi will be described as an example. In the filter unit, G data is input, and G data in which deterioration of the MTF characteristics is corrected by enhancing the edge of the image data is output.

  FIG. 5 shows the function of the document recognition unit 320 in block sections. The document recognition unit 320 performs character edge detection, pattern detection, and chromatic / achromatic detection, and outputs a C / P signal representing a character edge area or a pattern area and a B / C signal representing a chromatic area / achromatic area. appear.

  The document recognition unit 320 is roughly divided into a filter unit 321, an edge extraction unit 322, a white region extraction unit 323, a halftone dot extraction unit 324, a color determination unit 325, and an overall determination unit 326. Here, a case where the reading density of the scanner 11 is about 600 dpi will be described as an example.

  The filter unit 321 corrects the G image data generated by the scanner 11 mainly for extracting the edge of the character. Here, since the data read by the scanner 11 may be blurred due to the performance of a lens or the like, an edge enhancement filter is applied. However, here, it is necessary to simply emphasize the edge on the original and not to emphasize the line pattern for gradation expression that is widely used in copying machines. If the line pattern is emphasized, it is necessary to extract the picture (gradation expression area by the line pattern) as an edge and eventually misidentify it as a character edge. is there. Also, as shown in FIG. 9, the 600 dpi line pattern A and the 400 dpi line pattern B have different repetition periods, so it is difficult to avoid enhancement with the same filter coefficient. Therefore, the cycle of the image pattern is detected and the filter coefficient is switched. In FIG. 9, the sum of the width of one white block in the main scanning direction x and the width of one black block adjacent thereto is a line pitch (constant width: a predetermined number of pixels), that is, a line cycle. In this case, the white block width is widened and the black block width is narrowed. As the density becomes higher, the white block width becomes narrower and the black block width becomes wider.

  In the configuration example including the document recognition unit 320 in FIG. 5, the pixel matrix of the filter unit 321 is composed of the number of pixels in the main scanning direction x × the number of pixels in the sub-scanning direction y (the mechanical document scanning direction of the scanner 11). 5, as shown in the block of the filter unit 321 on FIG. 5, two sets of coefficients addressed to the respective weighting coefficients a1 to a7, b1 to b7, c1 to c7, d1 to d7, and e1 to e7. There are groups (coefficient matrices) A and B. The next coefficient group A is a filter processing coefficient that suppresses the enhancement of the 600 dpi line pattern A in FIG. 9 and emphasizes the edge of the character, and the coefficient group B is a 400 dpi line pattern in FIG. B is a coefficient for filtering that suppresses emphasis and emphasizes the edge of a character.

The filter process is a calculation result with the coefficient group A or B / 16 + the target pixel. This enhances the image data.
Coefficient group A0-10-20-100-100-20-20-100-1-120-1-100-10-20-20-100-10-20-10.

  Coefficient group B-100-200-1-100-200-1-10-120-10-1-100-200-1-100-200-1.

  Note that the horizontal direction is the alignment in the main scanning direction x, and the vertical direction is the alignment in the sub-scanning direction y. The coefficients in the first row of the coefficient groups A and B are the coefficients a1 to a7 in the first row of the coefficient matrix of the block of the filter 321 in FIG. The center “20” is the coefficient of the pixel in the center of the third row c1 to c7 of the coefficient matrix of the block of the filter unit 321, that is, the coefficient c4 of the target pixel. The sum (product sum value) of products (total 7 × 5 = 35) obtained by multiplying each coefficient of the coefficient matrix by the value represented by the image data of the pixel addressed to the coefficient is the pixel of interest (pixel addressed to c4). The image data values processed by the filter unit 321 are given to the edge extraction unit 322 and the white area extraction unit 323. Here, the pixel of interest is a pixel that is currently processed, and it is updated to pixels that are sequentially different in the x direction and in the y direction.

  In the coefficient group A, negative coefficients (small coefficient values) are distributed at the line pitch of the 600 dpi line pattern A shown in FIG. 9, and 0 (slightly large coefficient) is distributed between them. For emphasis, 20 (very large coefficient) is assigned to the target pixel. As a result, when the image data (target pixel) is the black / white edge of the line pattern A region, the weighted average value (product sum value) derived therefrom is the character edge that is not the line pattern A. Compared to a certain time, the value is considerably lower.

  In coefficient group B, negative coefficients (small value coefficients) are distributed at the line pitch of 400 dpi line pattern B shown in FIG. 9, and 0 (slightly large coefficient) is distributed between them, and For the edge enhancement, 20 (very large coefficient) is assigned to the target pixel. As a result, when the image data (target pixel) is the black / white edge of the line pattern B region, the weighted average value (product sum value) derived therefrom is the character edge that is not the line pattern B. Compared to a certain time, the value is considerably lower.

  Note that the filter unit 321 performs the filter processing by the coefficient group B when one of the following conditions 1 and 2 is satisfied, that is, when there is a high possibility that the line pattern B of 400 dpi in FIG. Sometimes filter processing by coefficient group A is performed.

<Condition 1> [Condition to see if the 400 dpi line pattern B is thin (white section in FIG. 9)]
(D [3] [1] <D [3] [2]) & (D [3] [7] <D [3] [6]) & (ABS (D [3] [2] -D [3 ] [4])
> ABS (D [3] [4] -D [3] [1])) & (ABS (D [3] [6] -D [3] [4])
> ABS (D [3] [4] -D [3] [7]))

<Condition 2> [Condition for checking whether the 400 dpi line pattern B is dark (black section in FIG. 9)]
(D [3] [1]> D [3] [2]) & (D [3] [7]> D [3] [6]) & (ABS (D [3] [2] -D [3 ] [4])
> ABS (D [3] [4] -D [3] [1])) & (ABS (D [3] [6] -D [3] [4])
> ABS (D [3] [4] -D [3] [7]))

  Note that D [i] [j] means a value represented by image data of a pixel at a position of x = i, y = j on a pixel matrix of x, y distribution, for example, D [3] [1 ] Is a value represented by image data of a pixel to which the coefficient a3 of the coefficient matrix shown in the block of the filter unit 321 in FIG. 5 is addressed. “&” Means “logical product: AND”, and “ABS” means an absolute value operator. The target pixel is D [4] [3].

  When the above condition 1 or 2 is satisfied, it is assumed that the pixel of interest at that time is in the 400 dpi line pattern pattern B area at the time of 600 dpi reading shown in FIG. Perform filtering. If neither of the conditions 1 and 2 is satisfied, the character edge emphasis filter process is performed using the coefficient group A that avoids emphasizing the 600 dpi line pattern A at the time of 600 dpi reading shown in FIG. That is, the image period (pitch) is detected so that the image pattern of a specific period is not emphasized. It is possible to emphasize the edge of a character without emphasizing the line pattern. FIG. 5 shows an aspect in which G image data is referred to for edge processing, but it is not limited to G image data, and may be luminance data. Any signal expressing dark and light can be applied.

The edge extraction unit 322 will be described.
The character area has many high-level density pixels and low-level density pixels (hereinafter referred to as black pixels and white pixels), and these black pixels and white pixels are continuous at the edge portion. The edge extraction unit 322 detects a character edge based on the continuity of each of such black pixels and white pixels.

The ternary unit 322a will be described.
First, the ternarization unit 322a uses the two thresholds TH1 and TH2 to trinify the G image data (input data of the edge extraction unit 322) that the filter unit 321 has performed the filter processing for character edge emphasis. For example, when the image data represents 256 gradations (0 = white) from 0 to 255, the thresholds TH1 and TH2 are set to TH1 = 20 and TH2 = 80, for example. The ternary unit 322a converts the input data into ternary data that represents the pixel to which the data is addressed as a white pixel if the input data <TH1, and the halftone pixel if TH1 ≦ input data <TH2. The input data is converted into ternary data to be expressed, and the input data is converted into ternary data to be expressed as a black pixel if TH2 ≦ input data.

The black pixel continuous detection unit 322b and the white pixel continuous detection unit 322c will be described.
The black pixel continuation detection unit 322b and the white pixel continuation detection unit 322c detect a location where the black pixels are continuous and a location where the white pixels are continuous based on the ternary data, respectively, by pattern matching. For the pattern matching, in the configuration example including the document recognition unit 320 of FIG. 5, the patterns BPa to BPd and WPa to WPd of the 3 × 3 pixel matrix shown in FIG. 10 are used. In the pattern shown in FIG. 10, black circles indicate the above-described black pixels, white circles indicate the above-described white pixels, and blank pixels without any circles are black pixels, halftone pixels, white pixels It does not matter whether it is any of the above. The pixel at the center of the 3 × 3 pixel matrix is the target pixel.

  When the distribution of the content of the ternary data matches one of the black pixel distribution patterns BPa to BPd shown in FIG. 10, the black pixel continuous detection unit 322b represents the target pixel at that time as “black continuous pixels”. Data is given to the pixel of interest. Similarly, when the white pixel continuous detection unit 322c matches any of the white pixel distribution patterns WPa to WPd shown in FIG. 10, the target pixel at that time is set as a “white continuous pixel” and data representing it is given to the target pixel. .

The neighboring pixel detection unit 322d will be described.
The next neighboring pixel detection unit 322d determines whether there are black continuous pixels or white continuous pixels in the vicinity of the target pixel in the neighboring pixel detection unit 322d with respect to the detection results of the black pixel continuous detection unit 322b and the white pixel continuous detection unit 322c. By checking whether or not the pixel of interest is in the edge region or the non-edge region, it is determined. More specifically, in the configuration example provided with the document recognizing unit 320 in FIG. 5, the block is a 5 × 5 pixel matrix, and one or more black continuous pixels and one or more white continuous pixels exist therein. Sometimes the block is determined to be an edge region, otherwise it is determined to be a non-edge region.

The isolated point removing unit 322e will be described.
Furthermore, since the character edge exists continuously, the isolated edge removing unit 322e corrects the isolated edge to a non-edge region. Then, an edge signal “1” (edge area) is output to a pixel determined as an edge area, and an edge signal “0” (non-edge area) is output to a pixel determined as a non-edge area.

  As shown in FIG. 19, the white area extraction unit 323 in FIG. 5 includes a binarization unit 323a, an RGB white extraction unit 323b, a white determination unit 323c, a white pattern matching unit 323d, a white pattern correction unit 323j, a white expansion unit 323h, It comprises a white contraction unit 323l, a white correction unit 323g, a gray pattern matching unit 323h, a gray expansion unit 323i, and a determination unit 323m. Note that the white region extraction unit 323 in FIG. 5 is obtained by replacing the portion M in FIG.

  The binarization unit 323a binarizes the edge emphasis output of the image density data (G image data) of the filter unit 321 with the threshold value thwsb, and the white pattern matching unit 323d (step SS7 in FIG. 6 representing the processing) A binary white determination signal for generating white data to be referred to is generated. The edge emphasis output is 256 gradations from 0 to 255 in the configuration example provided with the document recognition unit 320 of FIG. 5, 0 is white without density, and an example of the threshold value thwsb is 50. If the value of the edge enhancement output is smaller than thwsb = 50, the binarization unit 323a determines “binarized white” and generates a binarized white determination signal “1”. When the value of the edge emphasis output is thwsb = 50 or more, a binarized white determination signal “0” is generated.

  The RGB white extraction unit 323b performs 1) RGB white background detection, 2) color background detection, and 3) gray pixel detection to determine whether the image data is a white area or a gray area (medium density area).

1) RGB White Background Detection In the RGB white background detection, the white background separation operation is activated by detecting a white background area from R, G, B image data. That is, the white background separation process is started. Specifically, as shown in the pattern WBP of FIG. 11, if all of the R, G, B image data of the 3 × 3 pixel matrix is smaller than the threshold thwss, the target pixel (the central pixel of the 3 × 3 pixel matrix) is determined. The white area is determined and the white pattern matching unit 323d (white background determination signal referred to in step S3 in FIG. 11 representing the process) is activated (“1”). This is to detect whether there is a white pixel region of a certain extent. Each of the R, G, and B image data has 256 gradations from 0 to 255 in the configuration example provided with the document recognition unit 320 in FIG. 5, 0 is a base level without density, and threshold thwss < An example of thwsb, thwss, is 40, and if all of the R, G, and B image data are smaller than thwss = 40, it is determined as “white background” and a white background determination signal “1” is generated. When any of the R, G, and B image data is thwss = 40 or more, a white background determination signal “0” is generated.

2) Color Background Detection Color background is detected so that a light color is not determined as a white background.

A. Here, first, assuming that the sign of each pixel of the 5 × 5 pixel matrix centered on the pixel of interest is shown in the pattern MPp of FIG. 12, the center pixel c3 (× marked pixels of MCa to MCd) serving as the pixel of interest. When the RGB difference (difference between the maximum value and the minimum value of R, G, B image data addressed to one pixel) is larger than the threshold value thc, the color pixel determination signal a is set to “1” (color pixel), and is equal to or less than the threshold value thc. Is “0” (monochrome pixel).

B. If the R, G, B image data of any pixel in the peripheral pixel group Δ (in MCa to MCd in FIG. 12) on one side of the target pixel is all equal to or less than the threshold thwc, the one-side white determination signal b is set to “1”. ”(White pixel), and“ 0 ”(non-white pixel) when the threshold value thwc is exceeded. The threshold thwc is 20, for example.

C. If the R, G, B image data of any pixel in the peripheral pixel group □ (in MCa to MCd in FIG. 12) on the other side of the target pixel is all equal to or less than the threshold thwc, the other side white determination signal c is set to “1”. ”(White pixel), and“ 0 ”(non-white pixel) when the threshold value thwc is exceeded.

D. In any of the patterns MCa to MCd in FIG.
a AND (exclusive NOR of b and c) = "1"
That is, when a = “1” (the pixel of interest is a color pixel) and b and c match (a white pixel on both sides of the pixel of interest or a non-white pixel on both sides), The color background determination signal d is set to “1” (color background). This color ground determination signal d is referred to by the white pattern matching unit 323d (step S6 in FIG. 11 representing the processing).

  The above-described pattern matching A to D is performed because when the surroundings of the black characters are slightly colored due to the RGB reading position shift, they are not picked up as colors. At a colored position around the black character, (exclusive NOR of b and c) or “0” (one pixel on both sides of the target pixel is a white pixel and the other is a non-white pixel). In this case, the color gamut determination signal d = “0” (non-colored background). In addition, when the pixel of interest is a color pixel whose periphery is surrounded by a white background, the color background determination signal d = “1” (color background), and even when the line is intricate, a light color pixel is detected as the color background. be able to.

  That is, when the line is intricate, the originally white portion is not completely read in white. When the RGB difference is small, the pixel is not determined as a color pixel. Therefore, the threshold thwc is set to be stricter than the white background for which the density is to be viewed (for example, thws = 20 for thwss = 40 and thwsb = 50). In this process, it is possible to accurately check whether a white background is present and to accurately detect a light color pixel as a color background.

  In addition, in the case of color detection, valley white pixel detection may be performed. The valley white pixel detection detects valley white pixels in a small white area that cannot be detected by the above-described RGB white background detection based on the 5 × 5 pixel matrix distributions RDPa and RDPb of the G image data shown in FIG.

Specifically, based on the 5 × 5 pixel matrix distribution RDPa,
miny = min (G [1] [2], G [1] [3], G [1] [4], G [5] [2], G [5] [3], G [5] [4 ])
Is calculated. That is, the minimum density miny in the pixel group with black circles in the 5 × 5 pixel matrix distribution RDPa shown in FIG. 11 is extracted. And
maxy = max (G [3] [2], G [3] [3], G [3] [4])
Is calculated. That is, the maximum density maxy in the pixel group with white circles in the 5 × 5 pixel matrix distribution RDPa shown in FIG. 11 is extracted. next,
mint = min (G [2] [1], G [3] [1], G [4] [1], G [2] [5], G [3] [5], G [4] [5 ])
Is calculated. That is, the lowest density mint in the pixel group with black circles in another 5 × 5 pixel matrix distribution RDPb shown in FIG. 11 is extracted. And
maxt = max (G [2] [3], G [3] [3], G [4] [3])
Is calculated. That is, the maximum density maxt in the pixel group with white circles in the 5 × 5 pixel matrix distribution RDPb shown in FIG. 11 is extracted. Here, min () is a function for detecting the minimum value. max () is a function for detecting the maximum value. next,
OUT = ((miny-maxy)> 0) # ((mint-maxt)> 0)
Is calculated. That is, of (miny-maxy) and (mint-maxt), the larger positive value is defined as a valley detection value OUT, and if the value of OUT is equal to or greater than a certain threshold, the target pixel (RDPa or RDPb Are detected as valley white pixels.
In this way, the valley state of the image is detected, and 1) RGB white background detection compensates for difficult detection. Note that # means a logical sum (OR: or).

The white determination unit 323c will be described.
Here, the state variables MS and SS [I] used for white determination are updated. The contents are shown in the flowchart of FIG. Here, the state variable MS is addressed to the pixel of the processing target line (target line), and the state variable SS [I] is addressed to the pixel one line before the processing target line (processed line). This is 4-bit white background information representing the degree of white, and is generated by the processing shown in the flowchart of FIG. The maximum value represented by the state variables MS and SS [I] is set to 15, which means the whitest degree, and the minimum value is 0. That is, the state variables MS and SS [I] are data indicating the degree of white, and the larger the value represented, the stronger the white. At the start of the copying operation, the state variables MS and SS [I] are both initialized to 0.

  In the process of FIG. 6, first, the state variable of the target pixel to be processed one line before, that is, the white background information SS [I] and the pixel one pixel before the target pixel on the same line (previous pixel: processed pixel). The state variable, that is, the white background information MS is compared (step S1). If the white background information SS [I] one line before is larger, this is used as the temporary white background information MS of the target pixel (step S2). Otherwise, the state variable MS of the preceding pixel is set as the temporary white background information MS of the target pixel. This means that information closer to white of the white background information of the peripheral pixels is selected.

  When a white area, that is, a white background is detected by the above-described 1) RGB white background detection after starting the copying operation, [1) White background determination signal = “1”, which is the output of the RGB white background detection, the pixel one line before the target pixel The white background information SS [I] is updated to 15 (steps S3 and S4), and the white background information MS of the target pixel is also set to 15 (step S5). Then, the white background information MS of the target pixel is written in the main scanning position (F) of the target pixel in the line memory for the current line (target line) of the line memory LMP shown in FIG. SS [I] is written in the main scanning position (F) of the target pixel of the line memory for the previous one line of the line memory LMP shown in FIG. 13 (steps S3, 4, 5). Next, the white background information SS [I] addressed to the previous pixel is propagated to the previous pixel as follows (steps S14 to S17). [I] means the main scanning position of the target pixel, and [I-1] means the position of the pixel one pixel before (in the main scanning direction x) (the pixel immediately before the target pixel).

  When SS [I-1] <SS [I] -1, SS [I-1] = SS [I] -1 is set in the line memory (steps S14 and S15). That is, in the line one line before the target pixel, the white background at the position (F) of the target pixel from the white background information SS [I-1] one pixel before (E) from the position (F) of the target pixel in the main scanning direction. If the value “SS [I] −1” obtained by subtracting 1 from the information SS [I] is larger (whiteness is stronger), the pixel one pixel before the position (F) of the target pixel on the line one line before The white background information SS [I-1] addressed to (E) is updated to a value obtained by lowering the white intensity by 1 from the white background information SS [I] at the position (F) of the target pixel.

  Next, when SS [I-2] <SS [I] -2, SS [I-2] = SS [I] -2 is set in the line memory (steps S16, 17-14, 15); When SS [I-3] <SS [I] -3, SS [I-3] = SS [I] -3 is set in the line memory (steps S16, 17-14, 15).

  Similarly, finally, when SS [I-15] <SS [I] -15, SS [I-15] = SS [I] -15 is set in the line memory (steps S16, 17-14). 15). The lower limit value MIN of the value of the white background information SS [I] is 0, and when it is less than 0, it is kept at 0. The same applies to step S13 described later.

  By the processing of these steps 14 to 17, the white background information SS one line before and before the main scanning position of the target pixel is changed to one for the positional deviation of one pixel in the main scanning direction x. The value is updated to a value reduced by the reduction rate, and the white background information of the target pixel is propagated at the reduction rate in the main scanning direction x one line before the main scanning (white propagation processing). However, this is a case where the white background information one line before has a smaller value. For example, the pixel in the previous line is 1. ) When a white background (white area) is detected by RGB white background detection, the white background information is 15 and is the highest value, so rewriting is not performed.

  When the pixel of interest is updated and becomes a non-white background [1) White background determination signal = “0”, which is the output of RGB white background detection, the process proceeds from step S3 to step S6 and subsequent steps, and the pixel of interest changes to the color background [above. 2) Not a color gamut determination signal d = “1” which is an output of color gamut detection (a non-color gamut), binarized white [binary white determination signal which is an output of the binarization unit 323a] = “1”], and when the state variable of the target pixel tentatively determined in Steps 1 and 2, that is, the white background information MS is equal to or greater than the threshold thw1 (eg, 13), the white background information MS addressed to the target pixel is incremented by one. (Steps S6 to 10). That is, the white level is updated to a strong value by 1. The maximum value max of the white background information MS is set to 15, and when it exceeds 15, it is limited to 15 (steps S9 and S10). Even when the route is advanced, the above-described steps S5 and 14 to 17 are executed. That is, white propagation processing is performed.

  When the target pixel is non-colored and binary white, but the white background information MS is less than thw1 (for example, 7), thw2 (for example, 1) or more, and is a valley white pixel, the state variable MS is set to the value as it is. Hold (steps S8, 11, 12). Even when the route is advanced, the above-described steps S5 and 14 to 17 are executed. That is, white propagation processing is performed.

  If none of the above conditions is met, that is, if the target pixel is a color background or non-binary white, the white background information MS of the target pixel is decremented by 1 (step S13). In other words, the white level information is updated to white background information weak by one. The minimum value MIN of the white background information MS is 0, and is kept at 0 when it is less than 0. Even when the route is advanced, the above-described steps S5 and 14 to 17 are executed. That is, white propagation processing is performed.

  By generating the white background information MS as described above, the white information can be propagated to the peripheral pixels via the state variable (white background information) MS on the line memory LMP. As described above, the white background information MS is generated in step S3-4 in FIG. 6 based on the RGB white background determination signal that represents a white background when the color data (all of the R, G, and B image data) is smaller than the threshold thwss = 40. Including generation of white background information MS corresponding to colors of -5-14 to 17 and when the edge emphasis output of density data (G image data) (output of the filter unit 321) is smaller than the threshold thwsb = 50, This includes the generation of white background information MS corresponding to the density of the system in steps S7 to 13-5-14 to 17 in FIG. 6 based on the white background and the binarized white determination signal.

  The white determination unit 323c first detects 1) RGB white background in the RGB white extraction unit 323b until a white area is detected, that is, 1) RGB white background detection generates a white background determination signal “1” and responds to this. The operation (execution of step S4) is not performed until the generation of the color-corresponding white background information MS (steps S3-4-5-14 to 17) is started. This is to prevent a region that cannot be determined as a white region from being erroneously determined as a white pixel (white block) in white pattern matching (to be described later) of the G image data after edge enhancement processing by the filter unit 321.

  When the edge emphasis filter unit 321 is applied to a light color ground character, the data around the character becomes a value (white) having a lower level than the original image data (color ground), so the data after the edge emphasis processing of the filter unit 321 is performed. When white pattern matching is performed, that is, when white area determination is performed based only on generation of density-corresponding white background information MS (steps S7-13-5-14-17), the character ground around the color ground is erroneously determined as white. Although it is easy, white for determining a white pixel (white block) to be described later in an area that can be determined as a white area by generating the white background information MS corresponding to the color (steps S3-4-5-14 to 17). If the white background information MS is set to the highest value so that pattern matching is applied, and the white background is not white at step 3, the white background condition is checked in detail in step S6 and subsequent steps. Since the white background information MS, which is one parameter for determining whether to apply ching, is adjusted, white pixels (white blocks) are detected by white pattern matching (to be described later) of the G image data after edge enhancement processing by the filter unit 321. This prevents misjudgment.

  For example, when the possibility of a color pixel is high, the white background information MS is lowered (step S13), and when there is a suspicion of a color pixel, the white background information MS is held (no change) (steps S11 to 13). Pattern matching prevents erroneous determination as a white pixel (white block), and prevents data around the character from having a lower level (white) than the original image data (color background).

  When the character is dense, the white background information MS is updated and propagated by the above-described processing (steps S3 to 5, 6 to 10, and 14 to 17), so that the possibility that the dense character region is erroneously determined as a pattern is reduced. . In addition, among characters such as complicated characters (for example, “call”), there are cases where 1) white detection cannot be performed by RGB white background detection. In that case, 3) white is detected by valley white pixel detection, and white background is detected. Since the information MS is held by the route in which the YES output of step S12 goes straight to step S5 and remains in a white background tendency, the possibility that a complicated character is erroneously determined as a design is reduced.

  As described above, when the pixel of interest is a color pixel surrounded by a white background, the color gamut determination signal d = “1” (color gamut), which is the output of the above 2) color gamut detection, is obtained. Even in a complicated area, a light color pixel can be detected as a color background, and a threshold thwc for determining whether the periphery of the target pixel is white is set low (thwc = 20), and the periphery of the light color pixel (target pixel) is Since it is possible to detect a light color pixel as a color background by strictly checking whether the background is a white background, it is possible to further reduce the possibility that a complicated character is erroneously determined as a design.

  As described above, since a light color pixel can be detected more precisely as a color background, when a color background is detected, the process proceeds from step S6 to step S13 in FIG. 6, and the state variable MS is lowered to determine the color background as white. In addition to being able to reduce the possibility, in addition to the threshold thwss (eg, 40) when generating the white background determination signal referred to in step S3, the threshold when generating the binary white determination signal referred to in step S7 If thwsb (for example, 50) is set to a large value and the color background is not determined (step S6: NO), the probability that the binarization unit 323a regards it as white is increased, and steps S7 to S10 in FIG. The state variable MS is increased to increase the possibility of determining a white area.

  That is, in the above-described 1) RGB white background detection, the threshold value thwss = 40, and a strict white determination with a low probability of determining white is performed. When the white background is determined there, the state variable MS is obtained by the processing from step S3 to step 4 in FIG. To increase the possibility of judging the character background to be white. If the white background is not determined by the above strict white determination, conversely, a strict color gamut determination with high reliability for detecting a thin color pixel as a color gamut as a color gamut, that is, the above 2) color gamut detection. When the result is referred to and it is not determined to be a color background, once again, a sweet white determination with a threshold thwsb = 50 having a high probability of determining white, that is, referring to the binarization unit 323a, If the determination is white, the state variable MS is increased to increase the possibility that the character background is determined to be white (steps S7 to S10).

  Since the processing in steps S6 to S10 described above is performed, if the background density unevenness is thinner than the color background and the detected light color pixel, for example, if there is unevenness in the background of the document such as show-through, the fine background unevenness of the document. As a result, it is possible to prevent the state variable MS from significantly changing in binary, and to prevent the next white pattern matching unit 323d from determining whether or not the pixel is a white pixel in the scanning direction. As a result, when the background is a light color background, fine color loss (white background) does not appear in conjunction with the fine background unevenness of the document such as show-through.

The white pattern matching unit 323d will be described.
Whether the background is white is determined by whether or not there is a continuous white pixel in a 5 × 5 pixel unit block centered on the target pixel. Therefore, regarding the target pixel, when the following expression is satisfied, the target pixel is temporarily determined as a white pixel, and white pattern matching is performed according to the following conditional expression.
(Non-color pixel & (white background information MS ≧ thw1 (13)) & binarized white) # (Non-color pixel & (white background information MS ≧ thw2 (1)) & valley white pixel & binarized white)
Here, the target pixel for checking whether or not this conditional expression is satisfied is the target of the white propagation processing in steps S5 and 14 to 17 in FIG. “White background information MS” is the white background information MS [I] of the pixel of interest for which the above check after the white propagation process is performed. However, MS [I] is the white background information after the white propagation processing, and I is the position in the main scanning direction x of the target pixel to be checked, and the white determination unit 323c described above performs the state variable MS. This is different from the position in the main scanning direction x of the pixel of interest for which is calculated.

  In the conditional expression, “non-color pixel” indicates that the color gamut determination signal d, which is the output of 2) color gamut detection, is “0”, and “binary white” indicates that the binarization unit 323a The binarized white determination signal is “1” (binarized white), and “valley white pixel” means that the detection result of the above 3) valley white pixel detection is a valley white pixel. And # means logical OR (or :). The white pattern matching is for checking whether the output (whether it is a white pixel) determined by the above conditional expression corresponds to one of the vertical and horizontal diagonal continuity patterns PMPa to PMPd in FIG. White circles attached to the patterns PMPa to PMPd mean white pixels. It does not matter whether the other blank pixels are white pixels.

  If the white pixel distribution of the 5 × 5 pixel matrix centered on the target pixel corresponds to the pattern PMPa, PMPb, PMPc, or PMPd in FIG. 13, it is determined that the target pixel is a white pattern pixel.

  As gray pixel detection, hue division of R, G, B, Y, M, C, and Bk is performed, and a pixel having a low density is detected for each hue. Hue division is the same as color determination described later. Here, the filtered G data is compared with thgr, and if it is larger than the G data or if it is a color pixel in RGB white extraction color pixel detection, the following calculation is performed. If the condition is satisfied, the pixel is a gray pixel. Here, the threshold is changed for each color because the maximum density of each ink is different.

4.1) RY hue region boundary (ry)
R-2 * G + B> 04.2) Y-G hue region boundary (yg)
11 * R-8 * G-3 * B> 04.3) GC hue region boundary (gc)
1 * R-5 * G + 4 * B <04.4) CB hue region boundary (cb)
8 * R-14 * G + 6 * B <04.5) B-M hue area boundary (bm)
9 * R-2 * G-7 * B <04.6) MR hue region boundary (mr)
R + 5 * G-6 * B <04.8) Y pixel pixel determination (gry)
(It is a color pixel) & (ry == 1) & (yg == 0) & (Maximum RGB value <thmaxy)
4.9) G pixel determination (grg)
(It is a color pixel) & (yg == 1) & (gc == 0) & (Maximum RGB value <thmaxg)
4.10) C pixel determination (grc)
(It is a color pixel) & (gc == 1) & (cb == 0) & (Maximum RGB value <thmaxc)
4.11) B pixel determination (grb)
(It is a color pixel) & (cb == 1) & (bm == 0) & (Maximum RGB value <thmaxb)
4.12) M pixel determination (grm)
(It is a color pixel) & (bm == 1) & (mr == 0) & RGB maximum value <thmaxm)
4.13) R pixel determination (grr)
(It is a color pixel) & (mr == 1) & (ry == 0) & (Maximum RGB value <thmaxr)
4.14) When not a color pixel (grbk)
(RGB maximum value <thmaxbk)
4.15) Gray pixel determination When satisfying any of the conditions 4.8) to 4.15), a gray pixel is determined. Note that “==” is a C language notation.

  The above-described processing is performed by the gray pixel detection unit 323b-1 illustrated in FIG. As described above, the RGB white extraction unit 323b includes the gray pixel detection unit 323b-1, the color pixel detection unit 323b-2, and the RGB white background detection unit 323b-3, and R, G, and B image data are stored in these units. input. The output of the gray pixel detection unit 323b-1 is input to the gray pattern matching unit 323h, and the pattern matching result of the gray pattern matching is input to the gray expansion unit 323i. After performing expansion processing, the output is input to the determination unit 323m.

  In addition, the outputs of the color pixel detection unit 323b-2, the RGB white background detection unit 323b-3, and the binarization unit 323a are input to the white determination unit 323c, and the determination result of the white determination unit 323c is input to the white pattern matching unit 323d. The pattern matching result is input to the white pattern correction unit 323j and the white correction unit 323g. The correction result of the white pattern correction unit 323h is further processed by the white expansion unit 323k and the white contraction unit 323l and then input to the determination unit 323m, and the processing result of the white correction unit 323g is input to the determination unit 323m as it is. .

  It should be noted that the isolated points can be removed by performing a contraction process before the gray expansion unit 323i performs the expansion process. The white pattern matching unit 323d, the white pattern correction unit 323j, the white expansion unit 323k, the white contraction unit 323l, and the white correction unit 323g are configured to detect a boundary region that is not white and white, and are output from the white correction unit 323g. Indicates the line width, the output of the white contraction part 323l indicates a white region, and the output of the gray expansion part 323i indicates a medium density. Therefore, the determination unit 323m makes a determination by assigning priorities to these three outputs, and outputs the determination result to the subsequent stage. In this case, priority 1 is line width information from the white correction unit 323g, priority 2 is medium density information from the gray expansion unit 323i, and priority 3 is white area information from the white contraction unit 323l. is there.

  The gray pattern matching unit 323h performs the following pattern matching assuming that D is a gray pixel and bk is darker than the gray pixel. Since the copied original has a thin line pattern of 200 lines and a 300-line line, the following pattern is adopted so that the copied original is also detected in gray.

A pixel that matches the following pattern is a gray pixel. FIG. 20A and FIG. 20B show patterns at this time. FIG. 20A shows a pattern for 200 lines, and FIG. 20B shows a pattern for 300 lines.
(D15 & D25 & D35 & D32 & D45 & D38 &! BK41 & D42 &! BK43 &! BK44 & D55 &! BK46 &! BK47 & D48 &! BK49 D52 & D65 & D58 & D75 & D85 & D95) # (D05 & D15 & D25 & D31 & D33 & D35 & D37 & D38 & D41 &! BK42 & D43 &! BK44 & D45 &! BK46 & D47 &! BK48 & D48 && D51 & D53 & D55 & D57 & D58 & D65 & D75 & D85)

  The white pattern correction unit 323j deactivates active pixels that are isolated (white pixels of 1 × 1, 1 × 2, 2 × 1, 2 × 2, 1 × 3, and 3 × 1) by white pixel pattern matching. To do. This removes isolated pixels.

  In the white expansion unit 323k, the result of white pixel pattern matching correction is ORed by 7 × 41.

  The white contraction part 323l performs 1 × 33 AND of the result of white expansion in the white expansion part 323h. By performing white expansion and white contraction, inactive pixels existing in a small area with expansion are removed from the correction result in the white pixel pattern matching unit 323d. The determination result includes a boundary area on the non-white background side with respect to the white background and the boundary portion. In other words, the area is larger than the white background.

  In the white correction unit 323g, one or more white candidate blocks exist in each of the 6 × 4 pixel regions at the four corners in the 15 × 11 pixel centered on the target pixel marked with “×” in the block pattern BCP of FIG. The white block correction data is given to the target block. As a result, a region surrounded by a white background is set as a white region.

  The gray expansion unit 323i performs 11 × 11 OR processing on the gray pattern matching result. This makes the area slightly larger than the gray area.

The determination unit 323m sets the white background when the result of the white correction unit 323g is active or when the result of the white contraction unit 323l is active and the result of the gray expansion unit 323i is inactive. It can be expressed as the following equation.
White correction result # (white shrinkage result &! Gray expansion result)
Here, in the result of white correction, the area surrounded by the white background is definitely determined as the white area, and the result of white shrinking &! As a result of the gray expansion, a dark black character periphery is set as a white area, and a low density area is set as a non-white area.

The color determination unit 325 will be described.
When detecting color (chromatic) pixels and black (achromatic) pixels in a document, there is a relative reading shift of R, G, and B due to sampling of each color image data and mechanical accuracy. . This will be described with reference to FIG. FIG. 14A shows an image density signal. The black density signal is ideally black when the levels of the R, B, and G density signals coincide with each other. However, the actual image data is obtained by forming an image on the CCD 207 with the lens 206 and digitizing the image signal of the CCD 207. FIG. 14B shows an ideal high and low waveform. However, since a general scanner uses a three-line CCD sensor, the R, G, B line sensors are not read simultaneously in time, but the R, G, B line sensors are equal. Since they are arranged at intervals and cannot be read simultaneously in time, the reading position will inevitably shift. For example, the R, G, B color density signals representing the level-change black shown in FIG. 14B are relatively shifted as shown in FIG. 14C. If this shift is large, a color shift appears at the periphery of the black area.

The hue division unit 325a will be described.
The color determination unit 325 finds a chromatic color region. The input data R, G, B is converted into signals of c, m, y and color determination w (white) by the hue dividing unit 25a. As an example of hue division, each color boundary is obtained, and the difference between the maximum value and the minimum value of R, G, and B image data in one pixel is defined as an RGB difference, and the following is performed. Here, the R, G, B image data becomes black (darkens) as the number increases.

1) RY hue region boundary (ry) R-2 * G + B> 0
2) YG hue region boundary (yg) 11 * R-8 * G-3 * B> 0
3) GC hue region boundary (gc) 1 * R-5 * G + 4 * B <0
4) CB hue region boundary (cb) 8 * R-14 * G + 6 * B <0
5) BM hue region boundary (bm) 9 * R-2 * G-7 * B <0
6) MR hue region boundary (mr) R + 5 * G-6 * B <0
7) Color determination w (white) pixel determination: (R <thwa) & (G <thwa) & (B <thwa)
Then, y = m = c = 0. thwa is a threshold value.

8) Y pixel determination: (ry == 1) & (yg == 0) & (RGB difference> thy)
Then, y = 1 and m = c = 0. thy is a threshold value.

9) G pixel determination: (yg == 1) & (gc == 0) & (RGB difference> thg)
Then, c = y = 1 and m = 0. thg is a threshold value.

10) C pixel determination: (gc == 1) & (cb == 0) & (RGB difference> thc)
Then, c = 1 and m = y = 0. thc is a threshold value.

11) B pixel determination: (cb == 1) & (bm == 0) & (RGB difference> thb)
Then, m = c = 1 and y = 0. thb is a threshold value.

12) M pixel determination: (bm == 1) & (mr == 0) & (RGB difference> thm)
Then, m = 1 and y = c = 0. thm is a threshold value.

13) R pixel determination: (mr == 1) & (ry == 0) & (RGB difference> thr)
Then, y = m = 1 and c = 0. thr is a threshold value.

14) BK pixel determination: When not corresponding to 7) to 13), y = m = c = 1.

Further, the color determination w pixel is determined. condition is,
(R <thw) & (G <thw) & (B <thw)
Then, it is set as a w pixel for a color pixel and is output as w. thw is a threshold value. Here, the priority of 7) to 14) is given priority to the smaller number. The above-described threshold values thwa, thy, thm, thc, thr, thg, thb are threshold values determined before copying (processing). The relationship between thw and thwa is thw> tha. The output signal is 3 bits of data of 1 bit for each of c, m, and y, and 1 bit of w for color pixel detection for color determination. Here, the threshold is changed for each hue because the threshold corresponding to the hue area is determined for each hue area when the chromatic range is different. This hue division is an example, and any formula may be used.

  Output lines c, m, y, and w of the hue division unit 325a are stored in five lines in the line memories 325b to 325e and input to the color pixel determination unit 325f.

The color pixel determination unit 325f will be described.
FIG. 7 shows the contents of the color pixel determination unit 325f. The data of c, m, y, and w for five lines are input to the pattern matching units 325f5 to 325f7 and the count units 325f1 to 325f4. First, the pattern matching unit 325f6 in the flow for obtaining the B / C signal will be described.

The pattern matching unit 325f6 will be described.
When the color pixel w pixel exists, the pixel is corrected to c = m = y = 0. This correction increases the white level of the 5 × 5 pixel matrix centered on the pixel of interest. Next, the pixel of interest is a pixel other than all of c, m, y determined by the hue dividing unit 325a being 1 (c = m = y = 1) or all being 0 (c = m = y = 0) ( It is determined by checking whether the 5 × 5 pixel matrix matches the next pattern.

1) Color pixel pattern group 1-1) Pattern 1-1 (pm1) D23 & D33 & D43
1-2) Pattern 1-2 (pm2) D32 & D33 & D34
1-3) Pattern 1-3 (pm3) D22 & D33 & D44
1-4) Pattern 1-4 (pm4) D24 & D33 & D42

  The central pixel (target pixel) is D33. FIG. 15 shows these patterns pm1 to pm4. White circles on these patterns indicate that at least one of c, m, and y is 1. Pattern matching is used to prevent picking up isolated points. On the other hand, when detecting a small area color such as a halftone dot, the central pixel is a pixel (color pixel) other than 1 (c = m = y = 1) or all 0 (c = m = y = 0). It may be determined whether or not.

2) Detect a color line surrounded by a pattern group for color fine lines. The pattern used for this is shown in FIG. In FIG. 16, pixels with white circles are pixels in which c, m, and y are all 0. If the distribution of the data (c, m, y) of the 5 × 5 pixel matrix centered on the target pixel (center pixel) matches any of the patterns pw11a to pw14d in FIG. 16, the target pixel (center pixel) at that time ) Is regarded as a color line pixel.

2-1) Pattern 2-1 (pw11a to pw11d)
((D12 & D13 & D14) & (D42 & D43 & D44)) # ((D12 & D13 & D14) & (D52 & D53 & D54)) # ((D22 & D23 & D42) & (D42 & D43 & D44)) # ((D22 & D23 & D42) & (D52 & D53 & D54))
2-2) Pattern 2-2 (pw12a to pw12d)
((D21 & D31 & D41) & (D24 & D34 & D44)) # ((D21 & D31 & D41) & (D25 & D35 & D45)) # ((D22 & D23 & D24) & (D24 & D34 & D44)) # ((D22 & D23 & D24) & (D25 & D35 & D45))
2-3) Pattern 2-3 (pw13a to pw13d)
((D11 & D21 & D12) & (D35 & D44 & D53)) # ((D11 & D21 & D12) & (D45 & D44 & D55)) # ((D13 & D22 & D31) & (D35 & D44 & D53)) # ((D13 & D22 & D31) & (D45 & D44 & D55))
2-4) Pattern 2-4 (pw14a to pw14d)
(((D13 & D24 & D35) & (D41 & D51 & D52)) # ((D14 & D15 & D25) & (D41 & D51 & D52)) # ((D13 & D24 & D35) & (D31 & D42 & D53)) # ((D14 & D15 & D25) & (D31 & D42 & D53)

3) Perform pattern matching where all white area pattern groups c, m, and y are zero. The pattern used for this is shown in FIG. In FIG. 17, pixels with white circles are pixels in which c, m, and y are all 0. When the distribution of the data (c, m, y) of the 5 × 5 pixel matrix centered on the target pixel (center pixel) matches any of the patterns pw21a to pw24d in FIG. 17, the target pixel (center pixel) at that time ) Is regarded as a white area pixel.

3-1) Pattern 3-1 (pw21a to pw21d)
(D21 & D31 & D41) # (D22 & D32 & D42) # (D24 & D34 & D44) # (D25 & D35 & D45)
3-2) Pattern 3-2 (pw22a to pw22d)
(D12 & D13 & D14) # (D22 & D23 & D24) # (D42 & D43 & D44) # (D52 & D53 & D54)
3-3) Pattern 3-3 (pw23a to pw23d)
(D52 & D51 & D41) # (D53 & D42 & D31) # (D35 & D24 & D13) # (D25 & D15 & D14)
3-4) Pattern 3-4 (pw24a to pw24d)
(D54 & D55 & D45) # (D53 & D44 & D35) # (D31 & D22 & D13) # (D21 & D11 & D12)

4) Determination of color pixel candidate 2 If the pattern matching result extracted above matches the following pattern, the pixel of interest is set as color pixel candidate 2 for color determination: ((pm1 == 1) & ((pw11 = = 1) # (pw21! = 1))) # ((pm2 == 1) & ((pw12 == 1) # (pw22! = 1))) # ((pm3 == 1) & ((pw13 = = 1) # (pw23! = 1))) # ((pm4 == 1) & ((pw14 == 1) # (pw24! = 1)))

  Here, (pm1 == 1) means that the data distribution centered on the target pixel matches the pattern pm1, and (pw11 == 1) means that it matches any of the patterns pw11a to pw11d. This means that (pw21! = 1) matches with any of the patterns pw21a to pw21d. & Means logical product, and # means logical sum. With this pattern matching, a color pixel surrounded by a white area is set as a color pixel candidate, and when there is a white area other than that, it is not set as a color pixel. A color pixel candidate that matches in color pixel pattern matching without a white area is a color pixel candidate.

The count unit 325f1 will be described.
When there are w pixels for color determination in the 5 × 5 pixel matrix centered on the pixel of interest, the c, m, y data determined by the hue division unit 325a of the pixel is set to c = m = y = 0. to correct. This correction increases the white level of the pixel matrix. Then, the number of 1 of c, m, and y (c = 1, m = 1, y = 1) of each pixel in the pixel matrix is counted. If the difference between the maximum value and the minimum value of the count values for c, m, and y is greater than or equal to thcnt and the minimum value is less than thmin, color pixel candidate 1 is determined. thcnt and thmin are threshold values set before copying (processing). The plane is expanded to y, m, and c, the number is counted for each plane in the N × N matrix, and the minimum value is assumed to be black. This makes it possible to correct even if black pixel reading fails. A chromatic pixel is determined based on the difference between the maximum value and the minimum value. This corrects a pixel from which a black pixel is not read, and extracts a chromatic pixel. If there is a chromatic pixel of a certain pixel in a 5 × 5 pixel matrix centered on the pixel of interest, the pixel of interest is a chromatic pixel.

The color pixel determination unit 325f8 will be described.
Based on the outputs of the pattern matching unit 325f6 and the count unit 325f1, the color pixel determination unit 325f8 determines whether the pixel is a color pixel. If the color pixel candidate 1 and the color pixel candidate 2, the color pixel 1 is assumed.

The blocking unit 325f9 will be described.
The output of the color pixel determination unit 325f8 is blocked by the blocking unit 325f9. Blocking means that if there is one or more color pixels 1 in a 4 × 4 pixel matrix, the entire 4 × 4 pixel matrix is output as one color pixel block. In the processing after the block forming unit 325f9, 4 × 4 pixels are regarded as one block and output in block units.

The isolated point removing unit 325f10 will be described.
The isolated data is removed as an isolated point by the isolated point removing unit 325f10 if there is no color pixel block in the adjacent block of the target block.

The inflating part 325f11 will be described.
The output of the isolated point removing unit 325f10 is expanded to 5 × 5 blocks in the expansion unit 325f11 when there is one color pixel block. The reason for the expansion is to prevent black character processing around the color pixel. Here, the output B / C signal outputs L (chromatic) when the color pixel is one block, and outputs H (achromatic) at other times.

The count unit 325f2 will be described.
When a w determination pixel exists in the 5 × 5 pixel matrix centered on the target pixel, the c, m, y data determined by the hue division unit 325a of the pixel is corrected to c = m = y = 0. To do. This correction increases the white level of the pixel matrix. Then, the number of c, m, and y of 1 (c = 1, m = 1, y = 1) of each pixel in the pixel matrix is counted. If the difference between the maximum value and the minimum value of the count values for c, m, and y is greater than or equal to thacnt and the minimum value is less than thatmin, the pixel of interest is set as color pixel candidate 1. “thacnt” and “thamin” are threshold values set before copying (processing).

The color pixel determination unit 325f12 will be described.
Based on the outputs of the pattern matching unit 325f6 and the count unit 325f2, the color pixel determination unit 325f12 determines whether the pixel is a color pixel. If it is the color pixel candidate 1 and the color pixel candidate 2, it is set as the color pixel 2.

The blocking unit 325f13 will be described.
The output of the color pixel determination unit 325f12 is blocked by the blocking unit 325f13. That is, if there is one or more color pixels 2 in a 4 × 4 pixel matrix, the entire 4 × 4 pixel matrix is output as a color pixel 2 block. In the processing after the block forming unit 325f13, 4 × 4 pixels are set as one block and output in block units.

The density part 325f14 will be described.
In order to remove an isolated block, if there are three or more active conditions (color pixel 2 blocks) in the 3 × 3 block and the target block is active (color pixel), the target block is defined as an active block (color pixel 2 block). ).

The count unit 325f3 will be described.
The number of c, m, y of 1 (c = 1, m = 1, y = 1) of each pixel in the 5 × 5 pixel matrix centered on the target pixel is counted. If the difference between the maximum value and the minimum value of c, m, and y for each of c, m, and y is equal to or greater than tha1cnt, and the counted minimum value of c, m, and y is less than tha1min, color pixel candidate 3 is determined. tha1cnt and tha1min are threshold values set before copying (processing).

The pattern matching unit 325f5 will be described.
Whether the pixel (c, m, y) determined by the color pixel detection is a color pixel is determined by pattern matching using a 5 × 5 pixel matrix. The pattern is the same as that of the pattern matching unit 325f6. The pixel matched by pattern matching is set as a color pixel candidate 4.

The color pixel determination unit 325f15 will be described.
If it is the color pixel candidate 3 and the color pixel candidate 4, it is set as the color pixel 3.

The blocking unit 325f16 will be described.
The output of the color pixel determination unit 325f15 is blocked by the blocking unit 325f16. That is, if there is one or more color pixels 3 in a 4 × 4 pixel matrix, the entire 4 × 4 pixel matrix is output as a color pixel 3 block. The processing after the block forming unit 325f16 outputs 4 × 4 as one block.

The density part 325f17 will be described.
If there are three or more active conditions (color pixel 3 blocks) in the 3 × 3 block for removing the isolated block and the target block is active (color pixel 3), the target block is changed to the active block (color pixel 3). Block).

The count unit 325f4 will be described.
The number of 1 of c, m, and y (c = 1, m = 1, y = 1) determined by the hue dividing unit 325a of each pixel in the 5 × 5 pixel matrix centered on the target pixel is counted.

  If the minimum value of the count values of c, m, and y is greater than or equal to thabk, the pixel of interest is set as a black pixel candidate 1. thabk is a threshold value set before copying (processing).

The pattern matching unit 325f7 will be described.
In a 5 × 5 pixel matrix centered on the pixel of interest, pattern matching is performed on pixels with c = m = y = 1.

1-1) Pattern 1-1 (pm1) D23 & D33 & D43
1-2) Pattern 1-2 (pm2) D32 & D33 & d34
1-3) Pattern 1-3 (pm3) D22 & D33 & D44
1-4) Pattern 1-4 (pm4) D42 & D33 & D24
These patterns are shown in FIG. 15, and the pixels with circles in the figure are pixels with c = m = y = 1. When matching with any of these patterns, the target pixel is set as a black pixel candidate 2.

The achromatic determination unit 325f18 will be described.
If the pixel of interest is black pixel candidate 1 and black pixel candidate 2, it is determined as a black pixel.

The blocking unit 325f19 will be described.
The black pixel output is blocked by the blocking unit 325f19. In this block formation, if there is one or more black pixels in a 4 × 4 pixel matrix, the entire 4 × 4 pixel matrix is output as a black pixel block. The processing after the block forming unit 325f19 outputs 4 × 4 pixels as one block and outputs in blocks.

The expansion part 325f203 will be described.
If the block of interest is active (black pixel block) and its surrounding pixels are non-active (non-black pixel) in the × 3 block matrix, the block of interest is made non-active (non-black pixel block).

The general color pixel determination unit 325f21 will be described.
If the target block is determined to be active (color pixel 2) by the color pixel determination unit 325f12 and not determined to be active (black pixel) by the achromatic determination unit 325f18, the target block is determined to be a color (color block). Also, when the color pixel determination unit 325f15 is active (color pixel), the color is determined.

The inflating part 325f22 will be described.
In order for the integrated color pixel determination unit 325f21 to consider a small character as continuous for a block determined to be a color, if there is even one active block in the 9 × 9 block matrix centered on the target block, the target block Is the active block. Here, the reason for the large expansion is to fill the gap between the characters.

The continuous count unit 325f23 will be described.
The continuous counting unit 325f23 determines whether the original is a color document or a monochrome document by looking at the continuity of the color pixel blocks. By counting the number of continuous color pixels in the output data (color pixel block) of the expansion unit 325f22, it is determined whether the document is a color document.

  FIG. 8 shows the contents of this determination process. When the target pixel is in the color pixel block, the number of continuous color pixels of the target pixel is calculated with reference to the number of continuous color pixels of the upper left, upper, upper right, and left pixels of the target pixel (steps S21 to S26). Here, if the target pixel is, for example, the c3 pixel of the 5 × 5 pixel distribution pattern MPp in FIG. 12, the upper left, upper, upper right, and left pixels are the pixels b2, b3, b4, and c2, respectively. When the pixel of interest is not in the color pixel block, a continuous color pixel number of 0 is given to it (steps S21 to S27).

  When the target pixel is in the color pixel block, first, the number of continuous color pixels of the upper pixel (b3) of the target pixel (c3) is checked (step S22). A value obtained by adding 1 to the number of continuous color pixels of (b4) is given (step S24), and the number of continuous color pixels of the upper right pixel (b4) is added to the reference value A when the number of continuous color pixels of the upper pixel (b3) is 0. (Step S23). Next, a value obtained by adding 1 to the number of continuous color pixels of the upper left pixel (b2) is given to the reference value B, and a value obtained by adding 1 to the number of continuous color pixels of the upper pixel (b3) is given to the reference value B. Further, a value obtained by adding 1 to the number of continuous color pixels of the left pixel (c2) is given to the reference value D (step S25). Then, the highest value among the reference values A, B, C, and D is set as the number of consecutive color pixels of the target pixel (c3) (step S26).

  When the number of continuous color pixels is given to the target pixel (c3) as described above, it is checked whether the number of continuous color pixels is equal to or greater than the set value THACS (step S28). Is determined (step S29), and the processing of the continuous count unit 325f23 is ended there. If the number of continuous color pixels is less than the set value THACS, the target pixel is updated to the next pixel in the scanning directions x and y, and the above-described processing is repeated. As a result of performing the above-described processing on the entire surface of the document, if the number of continuous color pixels is less than the set value THACS until the end (steps S30 to S34), it is determined that the document is a monochrome image.

  The number of continuous color pixels is substantially the sum of vertical colored line segments and horizontal colored line segments. The number of consecutive color pixels in the upper right is different from others because it prevents double counting. Specific data on the number of continuous color pixels is shown in FIG. A small square including a number shown in FIG. 18 is a color pixel, and a number is a continuous number of color pixels given to the pixel. A block consisting of a series of small squares with numbers is a color pixel group, and even if one continuous color pixel in any color pixel group on the same document exceeds the set value THACS, it is a color document. The determination of color or black and white is confirmed (steps S28 and 29).

  The reason why it is divided from the color pixel determination units 1 to 3 (325f8 to 325f15) is to increase the accuracy of determining whether a color document or a monochrome document. The color pixel determination for black character processing is local and not so noticeable even if an erroneous determination is made. However, the determination of whether the original is a color original or a monochrome original affects the entire original if an erroneous determination is made. Therefore, the counting units 325f1-f4 are independent. Originally, it is better to make the hue dividing unit 325a independent. However, making the hue dividing unit 325a independent is not preferable because the memory of the pattern matching units 325f5-f7 increases. The increase in the amount of memory is reduced by changing the color pixel parameter (color pixel 1-3) with the parameters (color pixel candidate 1, 3 and black pixel candidate 1) of the count unit 325f1-f4. . The color pixel determination units 2 and 3 (325f12 and 325f15) are provided to detect a low density color such as yellow of a fluorescent pen, and further include an achromatic determination (black pixel determination) unit 325f18. This is because correction is made in the case of erroneous detection if the density is lowered. Light colors such as highlighters can be corrected with black data to a certain extent without any problem. When extracting multiple color pixels, only the level of w (white) is changed, so there is no need to have two memories for color pixel detection, and a capacity of one + 1 line is possible. is there.

  Since the count value is counted by the continuous count unit 325f23 by referring to the count data of the previous line and the count data of the current line, it is possible to accurately count the continuity of the peripheral pixels. It is possible to count continuations. In the configuration example including the document recognition unit 320 in FIG. 5, the hue determination is performed on the R, G, and B image data, but the present invention is not limited to the R, G, and B image data. ) Etc., it is easy to determine the hue.

As described above, the filter unit 251, the white region extraction unit 253, the edge extraction unit 252, and the color determination unit 254 of the image area separation unit 125 according to the embodiment of the present invention include the document recognition unit 320 of FIG. 5 described above. The filter unit 321, the white region extraction unit 323, the edge extraction unit 322, and the color determination unit 325 in the configuration example are configured in the same manner.
However, the white region extraction unit 253 according to the present embodiment outputs a binary signal wh = 1 when the target pixel is determined to be a white region, and outputs a binary signal wh = 0 when it is determined that the target pixel is not a white region. . The final white background separation result wh includes a boundary region on the non-white background side with respect to the white background and the boundary portion. In other words, the area is larger than the actual white background on the document.
The color determination unit 254 according to the present embodiment is a binary signal when the color determination unit 325 in the configuration example including the document recognition unit 320 in FIG. 5 described above determines that the pixel of interest is a chromatic pixel. When iro = 1 is output and it is determined that the pixel is an achromatic pixel, a binary signal iro = 0 is output.

Next, the halftone dot separation unit 255 in the present embodiment will be described with reference to FIG.
The halftone dot separation unit 255 includes a first halftone dot peak detection unit 261, a second halftone dot peak detection unit 262, a halftone dot region detection unit 263, and a halftone dot total determination unit 264.
G image data and B image data are input to the first halftone peak detector 261 and the second halftone peak detector 262. The G image data is image data that is sensitive (high sensitivity) to the black density of the color image, and it is not necessary to convert the image data into a luminance signal or the like. However, the G image data is not sensitive to Y (yellow: yellow), and the peak detection based on it is a leak of the Y peak. Since the B image data is sensitive to Y, the peak detection based on it can compensate for the leakage of the peak detection using the G signal. Therefore, the accuracy of halftone dot detection is increased.

The first halftone peak detector 261 outputs a binary signal gpk as a detection result, and the second halftone peak detector 262 outputs bpk. The halftone dot region detection unit 263 receives gpk, bpk, and the chromatic pixel detection result iro of the color determination unit, and generates a binary signal ami as the halftone dot region detection result.
As will be described later, since the handling of gpk and bpk varies depending on the value of iro, the halftone detection result ami also varies. Ami is input to the halftone dot total determination unit 264, and the final halftone dot separation result ht is output.

  The first halftone dot peak detection unit 261 uses a pixel density information in a two-dimensional local region having a predetermined size using image data to form a pixel that forms part of a halftone dot (referred to as a halftone dot peak pixel). To detect. When <Condition 1> and <Condition 2> shown below are satisfied simultaneously for the local region, the center pixel of the region is detected as a halftone dot pixel, and gpk = 1 is output. Otherwise, gpk = 0 is output.

<Condition 1> The density level of the central pixel is maximum (mountain peak) or minimum (valley peak) in the local region.
<Condition 2> The absolute value of the difference between the average density level of the pixel pair and the density level of the central pixel is equal to or greater than the threshold value Th for all pixel pairs that are point-symmetric with respect to the central pixel.

  With reference to FIG. 22 and FIG. 23, the detection process of the 1st halftone peak detection part 261 is demonstrated concretely. An M × M pixel matrix as shown in FIG. 22 is adopted as the local region. In the present embodiment, the case of M = 5 will be described, but this is not limited to the case of M = 5.

If the sign of each pixel of the M × M pixel matrix is as shown in the pattern of FIG. 23, the density Lc of the center pixel c3 that is the target pixel is the surrounding pixel (indicated by squares in MPa and MPb of FIG. 23). And the maximum or minimum compared to the concentrations L1 to L8 of
abs (2Lc-L1-L8) ≥Th and abs (2Lc-L2-L7) ≥Th and abs (2Lc-L3-L6) ≥Th and abs (2Lc-L4-L5) ≥Th
At this time, the center pixel (Lc) is detected as a halftone dot peak pixel, and a halftone dot detection signal gpk = 1 is generated. Otherwise, gpk = 0 is generated. Here, the abs function means taking the absolute value of the argument. Th is a threshold value (fixed value).

  Specifically, when one of the halftone peak pixel detections based on the peripheral pixel distribution patterns MPa and MPb shown in FIG. 23 is detected as a halftone peak pixel, the target pixel (center pixel c3) at that time is halftone. A point peak detection signal pk = 1 is given. The reason for using the two patterns is to deal with a wide range of halftone dot lines.

The pattern MPa is
L1 = b2, L2 = b3, L3 = b4, L4 = c2, L5 = c4, L6 = d2, L7 = d3, L8 = d4,
It is determined. Here, L1 = b2 means that the density of the pixel b2 is set to the value of L1 in the above-described halftone peak pixel detection calculation.

The pattern MPb is
L1 = b2, L2 = a3, L3 = b4, L4 = c1, L5 = c5, L6 = d2, L7 = e3, L8 = d4,
It is determined.

  When operating as a copying machine, reduction or enlargement in the sub-scanning direction is adjusted according to the size of the document scanning speed of the scanner 11. At this time, an image reduced or enlarged in the sub-scanning direction is input from the scanner 11 to the scanner correction unit 12. This case can be dealt with by changing the above two patterns. Specifically, the pattern MPc and the pattern MPd are used when the reduced image is input, and the pattern MPe and the pattern MPf are used when the enlarged image is input. As the peripheral pixel, a pixel displayed in a rectangle in the figure is used, but when inputting an enlarged image, a pixel displayed in a triangle may be added to the peripheral pixel.

  The second halftone dot peak detection unit 262 receives the B signal instead of the G signal input by the first halftone peak detection unit 261. The internal peak detection process is exactly the same as the first halftone peak detection unit 261. However, when the center pixel of the region is detected as a halftone dot peak pixel, bpk = 1 is output. Otherwise, bpk = 0 is output.

  The halftone dot area detection unit 263 receives gpk, bpk, and iro. The dot area detection processing by the dot area detection unit 263 will be described with reference to the flowchart of FIG.

[1] When iro = 1, that is, when the target pixel is determined to be chromatic (step S41)
In this case, a pixel with gpk = 1 or bpk = 1 (logical sum of gpk and bpk) is regarded as a halftone dot pixel, and the number of pixels is counted for each two-dimensional small region having a predetermined size. The sum of the halftone dot peak pixels of the valley is defined as a count value P. When the count value P is larger than the threshold value Pth, all the pixels in the two-dimensional small area (or in the case of pixel unit processing, only the central pixel of the small area) are determined as halftone dot areas. As a result, when it is determined as a halftone dot region, ami = 1 is output. Otherwise, output ami = 0. In the case of [1], by taking the logical sum of gpk and bpk, a halftone dot (for example, a yellow halftone dot) having a hue insensitive to the G signal sensitive to the black density of the image data can be detected as a halftone dot region. .

[2] When iro = 0, that is, when the target pixel is determined to be achromatic (step S42)
In this case, a pixel with gpk = 1 and bpk = 1 (logical product of gpk and bpk) is regarded as a halftone dot pixel, and the number of pixels is counted for each two-dimensional small region having a predetermined size. The sum of the halftone dot peak pixels of the valley is defined as a count value P. When the count value P is larger than the threshold value Pth, all the pixels in the two-dimensional small area (or in the case of pixel unit processing, only the central pixel of the small area) are determined as halftone dot areas. As a result, when it is determined as a halftone dot region, ami = 1 is output. Otherwise, output ami = 0. In the case of [2], since the pixel of interest is an achromatic pixel, a halftone dot region can be originally detected only by the G signal sensitive to the black density of the image data. In the prior art, the logical sum of gpk and bpk has always been used as a halftone dot pixel in order to increase the sensitivity to yellow halftone dots, but this has the disadvantage that black characters are easily misidentified as halftone dots. . In particular, when the influence of the RGB reading position shift of the scanner 11 is large, the number of halftone dots in the character area increases, and as a result, the probability that the character is erroneously determined as the halftone area increases. Here, in the case of [2], the AND of gpk and bpk is the halftone dot peak pixel. However, when the reading position deviation of the scanner RGB is large, the halftone dot detection performance may be lowered. At this time, halftone dot region detection may be performed by regarding only gbk as halftone dot peak pixels.

  In this embodiment, the probability of erroneously determining a black character as a halftone dot is reduced, but the probability of erroneously determining a color character as a halftone dot is not reduced. However, in general, it is recognized that the deterioration of image quality is greater when black characters cannot be determined due to erroneous determination of black characters. Therefore, it can be said that it is important to reduce the probability of erroneously determining a black character.

  The halftone dot region detection result ami detected in [1] and [2] as described above is input to the halftone dot total determination unit 264. The halftone dot peak pixels actually in the halftone dot area of the original are detected as a cluster having a certain size, but when isolated pixels with gpk = 1 or bpk = 1 are detected, There is a high probability that it is caused by noise or mere image noise. Therefore, by increasing the value of the threshold value Pth, the probability that an isolated pixel having gpk = 1 or an isolated pixel having bpk = 1 is determined as a halftone dot region can be reduced.

  The halftone dot comprehensive determination unit 264 receives the detection result ami of the halftone dot region detection unit 263. In a two-dimensional small region of a predetermined size of interest, the number of pixels with ami = 1 is counted, and when the total value AmiP is larger than a predetermined threshold value Amith, a halftone dot total determination unit 264 Determines that the target two-dimensional small region (target small region) is finally in the halftone dot region, and outputs a halftone dot detection signal ht = 1. Otherwise, it is determined as a non-halftone area and a halftone detection signal ht = 0 is output. The threshold value Amith is changed according to feature information in the vicinity of the target small region. That is, in the vicinity of the target small region, the threshold value Amith is changed in accordance with the result ht output from the halftone dot total determination unit 264 in the region processed before the target small region (processed region).

  In the present embodiment, two values TH1 and TH2 (TH1> TH2) are prepared as the threshold value Amith, and the halftone dot comprehensive determination of the processed region in the vicinity of the small region of interest input to the halftone dot total determination unit 264 One value is selected according to the result ht. That is, when the processed area in the vicinity is often determined to be a non-halftone area (ht = 0), the small area of interest is likely not to be a halftone area, so that false detection is reduced. The threshold value Amith is selected as TH1 where the condition becomes severe. On the other hand, if it is often determined that the nearby processed area is a halftone dot area (ht = 1), the target small area is likely to be a halftone dot area. TH2 for which the value becomes loose is used as the threshold value Amith. It should be noted that when image data is first input to the halftone dot total determination unit 264, there is no pixel determined as ht = 1. Therefore, TH1 is selected as the initial value of the threshold value Amith. Therefore, a comparison is made between the total number AmiP of pixels with ami = 1 in a predetermined region of interest and the threshold value Amith (= TH1).

  A specific example will be described with reference to FIG. FIG. 25 shows the distribution of the above-described small regions (two-dimensional small regions of a predetermined size), and each of T1 to T9 in the small region distribution pattern is a small region to which T5 is focused, T1, T2, T3. And T4 are processed small areas. When it is determined that ht = 1 for all the regions T1, T2, T3, and T4, TH2 (for example, 4) is selected as the threshold value Amith for determining the halftone dot region of T5. If any one of T1, T2, T3, and T4 is determined to be ht = 0, TH1 (for example, 7) is selected as the threshold value Pth. However, this is an example, and TH2 is selected when any one of T1, T2, T3, and T4 is determined to be a halftone dot region, and only when all are determined to be non-halftone dot regions. You may make it select TH1. Furthermore, the neighborhood area to be referred to when selecting the threshold value can be T1 only or T2.

The overall determination unit 256 includes a character determination unit 271, an expansion processing unit 272, and a decoding unit 273 as shown in FIG.
In the character determination unit 271, when the document area is edge-detected, white background is determined, and halftone dot is not determined, that is, when edge = 1, wt = 1, and ht = 0, the document area is at the character edge. Is determined. If the character edge is determined, the binary signal moji = 1 is output, otherwise moji = 0 is output. The determination result moji is input to the expansion processing unit 272.

  In the expansion processing unit 272, the result of the character determination unit 271 is subjected to OR processing of 8 × 8 blocks, and then AND processing of 3 × 3 blocks is performed to perform expansion processing of 4 blocks. That is, if any block of the 8 × 8 block centered on the target block is a character edge, it is assumed that the target block is also a character edge block, and all 3 × 3 blocks centered on the target block are If it is a character edge, the target block is determined as the character edge, and the target block and three blocks adjacent to the target block are regarded as the character edge. The reason for performing the AND process after the OR process is that, particularly in the case of a black character, if a small non-black character area exists around the black character area, a sense of incongruity may be felt due to a difference in processing. For example, black may appear light. In order to prevent this, the non-black character area is enlarged by OR processing. The AND process is performed to obtain a desired expansion amount.

  The C / P signal finally output by the decoding unit 273 is as shown in the following table. In addition, when C / P signal = 2, it does not exist in this embodiment.

  As for iro which is a signal indicating the color determination result, B / C = 1 is output if iro = 1, and B / C = 0 is output if iro = 0.

  Next, description will be made with reference to FIGS. 2 and 3 again. The C / P signal and B / C signal generated by the image area separation unit are converted into image data by the filter processing unit 122, the color correction unit 123, the scaling processing unit 124, the printer γ correction unit 181 and the gradation processing unit 182. Synchronously given to the cascade. The filter processing unit 122 is a filter that corrects the MTF of RGB data, and includes a coefficient matrix corresponding to an N × N pixel matrix, and logic that multiplies each coefficient by each image data to obtain a weighted average value. . When the C / P signal represents 1 (character area), a coefficient matrix for sharpening processing is used, and when it represents 0 or 3 (design area, halftone dot area), coefficients for smoothing processing A weighted average value is derived using the matrix and is output to the color correction unit 123. Here, the smoothing filter has a smoothing amount that increases in the order of the picture area and the halftone dot area. This is because a halftone dot structure remains unless the smoothness of the halftone dot is increased, resulting in moire.

  The color correction unit 123 converts the R, G, and B data into C, M, and Y data by a primary masking process or the like, and further performs UCR processing. By this UCR processing, color reproduction of image data can be improved. In the UCR process, the common part of the C, M, Y data is subjected to under color removal processing to generate Bk data, and the color correction unit 123 thereby outputs the C, M, Y, Bk data. Here, when the C / P signal is 1 (character area), full black processing is performed. Further, when the C / P signal is 1 and the B / C signal is 0 (achromatic region), the C, M, and Y data are erased. This is because black characters are expressed only by the black component.

  The printer γ correction unit 181 performs processing to change the γ curve according to the frequency characteristics of the plotter 19 and the C / P signal. When the C / P signal is 0 or 3 (picture area or halftone dot area), a γ curve that faithfully reproduces the image is used. When the C / P signal is 1 (character area), the γ curve is set up. To emphasize contrast.

  The gradation processing unit 182 performs quantization such as dither processing and error diffusion processing according to the gradation characteristics of the plotter 19 and the C / P signal. During Bk image formation, gradation-oriented processing is performed when the C / P signal is 0 or 3 (picture area or halftone area), and resolution is achieved when the C / P signal is 1 (character area). Perform important processing. For image formation other than Bk, gradation-oriented processing is performed when the C / P signal is 0 (picture area), and resolution-oriented processing is performed otherwise. The image data of each color “C” M ”Y” K ”subjected to the above processing is given from the printer correction unit 18 to the plotter 19 in synchronization with the image data writing operation.

  As described above, according to the above-described embodiment of the present invention, it is possible to accurately separate a character area and a halftone dot area in particular with black characters, so that image quality is improved by switching image processing between the two. be able to.

In the above-described embodiment of the present invention, the original image reading means, the chromatic area detecting means for detecting a chromatic area from the color image data acquired by the reading means, and the two-dimensional image data acquired by the reading means. A comparison determination unit that determines whether or not the absolute value of the density difference between the target pixel and the surrounding pixel group is larger than a preset threshold value with reference to the small area, and the comparison determination unit A halftone dot pixel detection unit having the pixel of interest determined to be larger than the threshold as a halftone dot peak pixel, a halftone dot peak pixel detected by the halftone dot pixel detection unit, and a halftone dot at the periphery thereof A halftone dot area detecting means for detecting a neighboring pixel including a target small area of the image data as a halftone dot area from the relationship with the halftone dot peak pixel detected by the peak pixel detecting means; and The dot region detecting means, depending on the chromatic region detection result, so that changing the dot region detection result.
For this reason, according to the above-described embodiment, the probability of erroneously determining a small character in document image data as a halftone dot can be reduced by reducing the probability of erroneously determining a black character rather than a color character, thereby improving black character quality.

Further, in the above-described embodiment of the present invention, among the color image data acquired by the reading unit, the first image data sensitive to the black density of the color image and the color component having low sensitivity of the first image data are sensitive. And a halftone dot area detecting unit for detecting a halftone dot area from the color image data using at least two pieces of image data of certain second image data.
For this reason, according to the above-described embodiment, it is possible to improve the quality of black characters while using a signal such that all colors in the document image data are sensitive.

  In addition, since the first image data is G image data, signal conversion processing from input image data to a luminance signal or the like is performed by using the G signal as data having high sensitivity to black density in the document image data. The need to do so can be eliminated.

  In addition, since the second image data is B image data, by using this B signal that is sensitive to the Y signal, which is less sensitive to the luminance signal, more accurate than halftone detection using only the luminance signal. Black character quality can be improved while performing high halftone dot detection.

Further, in the above-described embodiment of the present invention, the logical sum of the halftone dot detection result of the first image data and the halftone peak detection result of the second image data in the region chromatically determined by the chromatic region detection result. When the halftone dot detection unit determines that the target pixel is a halftone dot region, the first image is determined in the region that is determined to be a halftone dot region and determined achromatic by the chromatic region detection result. If the halftone dot detection unit determines that the target pixel is a halftone dot region using the logical product of the halftone dot detection result of the data and the halftone dot detection result of the second image data, the target pixel is determined to be a halftone dot region. Like to do.
Therefore, according to the above-described embodiment, the probability that a black character is erroneously determined as a halftone dot and the probability that a color character is erroneously determined as a halftone dot are changed by changing the reference for halftone dot region detection based on the chromatic determination. By lowering the value, the quality of black characters can be improved.

Further, the above-described effect obtained by the present embodiment is that a storage medium (or recording medium) in which a program code of software having the function of the above-described embodiment is recorded is supplied to the system or apparatus, and the computer of the system or apparatus Can also be obtained by reading and executing the program code stored in the storage medium. In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the storage medium storing the program code constitutes the present invention.
As this recording medium, for example, a floppy (registered trademark) disk, hard disk, optical disk, magneto-optical disk, CD-ROM, CD-R, magnetic tape, nonvolatile memory card, ROM, EEPROM, or the like may be used.
Further, by executing the program code read out by the computer, not only the functions of the above-described embodiments are realized, but also an operating system running on the computer based on the instruction of the program code performs actual processing. In some cases, the functions of the above-described embodiments are realized by performing part or all of the above.
Furthermore, after the program code read from the storage medium is written into a memory provided in a function expansion card inserted into the computer or a function expansion unit connected to the computer, the function is determined based on the instruction of the program code. This includes a case where the CPU or the like provided in the expansion card or the function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing.

  As the recording medium on which the program according to the present invention is recorded, for example, a recording medium on which an image processing program code is recorded, the program code includes an image input process code for inputting an image signal, and the image input A chromatic area detecting step for detecting a chromatic area from the color image data input in the step, and an absolute value of the density difference between the pixel of interest and the density of surrounding pixels in the predetermined area of the image input in the image input step. Each of which is determined to be larger than a preset threshold value and the pixel of interest whose absolute value of the density difference is determined to be larger than the threshold value by the comparison and determination step. A halftone dot peak detection step for determining a pixel as a peak pixel, a halftone dot peak pixel detected by the halftone dot peak detection step, and a neighborhood including a small area of interest. A halftone dot region detecting step for detecting a neighboring pixel including a target small region of the image data as a halftone dot region from the relationship with the halftone dot peak pixel, and in the halftone dot region detecting step, Depending on the result of the chromatic area detection process, the result of the halftone area detection process may be changed.

  Of the color image data acquired by the image input step, the first image data sensitive to the black density of the color image, and the second image data sensitive to the color component having low sensitivity of the first image data, And a halftone dot region detecting step of detecting a halftone dot region from the color image data using at least two pieces of image data.

  Further, in the area determined to be chromatic by the chromatic area detection step, the halftone dot area is obtained by using a logical sum of the halftone dot peak detection result of the first image data and the halftone peak detection result of the second image data. When the detection step determines that the area is a halftone dot area, the target pixel may be determined as a halftone dot area.

  Further, in the region determined to be achromatic by the chromatic region detection step, the halftone dot region is obtained by using a logical product of the halftone peak detection result of the first image data and the halftone peak detection result of the second image data. When the detection step determines that the area is a halftone dot area, the target pixel may be determined as a halftone dot area.

  According to the program according to the present invention and the recording medium on which the program is recorded, each function in the image processing apparatus as each embodiment according to the present invention described above can be realized in an apparatus controlled by the program. it can.

1 is a block diagram illustrating a configuration of an image processing apparatus as an embodiment of the present invention. 2 is a block diagram illustrating a configuration of a scanner correction unit 12 in the image processing apparatus. FIG. 2 is a block diagram illustrating a configuration of a printer correction unit 18 in the image processing apparatus. FIG. 3 is a block diagram illustrating a configuration of an image area separation unit 125 in the scanner correction unit 12. FIG. 3 is a functional block diagram illustrating a document recognition unit 320 in a configuration example including a document recognition unit 320. FIG. It is a flowchart which shows the process sequence of update of state variables MS and SS [I] used for the white determination in this structural example. It is a functional block diagram which shows the content of the color pixel determination part in this structural example. It is a flowchart which shows the process sequence of the determination process which determines whether it is a color original or a monochrome original in the continuous count part in this structural example. It is a figure which compares and compares the 600 dpi line pattern and the 400 dpi line pattern in this structural example. It is a figure which shows the pattern of a 3x3 pixel matrix used for the pattern matching performed in the black pixel continuous detection part in this structural example, and a white pixel continuous detection part. It is a figure which shows the example of a pattern used for the white background isolation | separation of the RGB white background detection part in this structural example. It is a figure which shows the example of a pattern used at the time of the color ground detection in this structural example. It is a figure which shows typically the present line (target line) of the line memory LMP in this structural example. It is a figure for demonstrating the process of the color determination part in this structural example. It is a figure for demonstrating the pattern matching in the color pixel determination in this structural example. It is a figure which shows the pattern for color thin lines which detects the color line surrounded by white in this structural example. It is a figure which shows the white area | region pattern for performing the pattern matching in which c, m, and y are all 0 in this structural example. It is a figure which shows the specific data of the color pixel continuous number in this structural example. It is a block diagram which shows the detail of the white area | region extraction part in this structural example. It is a figure which shows the example of a pattern matching with the line pattern for detecting the gray pixel in this structural example, (a) is a pattern for 200 lines, (b) is a pattern for 300 lines . FIG. 5 is a block diagram illustrating a configuration of a halftone dot separation unit 255 in the image area separation unit 125 in FIG. 4. 6 is a diagram illustrating an example of an M × M pixel matrix used by a first halftone peak detection unit 261. FIG. It is a figure which shows the example of a specific pattern of this MxM pixel matrix. 10 is a flowchart showing a halftone dot region detection process by a halftone dot region detector 263. It is a figure which shows the example of a small area distribution pattern. FIG. 5 is a block diagram illustrating a configuration of a comprehensive determination unit 256 in the image area separation unit 125 of FIG. 4.

Explanation of symbols

11 Scanner 12 Scanner Correction Unit 13 Compression Processing Unit 14 Controller 15 HDD
16 NIC
Reference Signs List 17 Decompression processing unit 18 Printer correction unit 19 Plotter 121 Scanner γ correction unit 122 Filter processing unit 123 Color correction unit 124 Scaling processing unit 125 Image area separation unit 181 Printer γ correction unit 182 Tone processing unit 251 Filter unit 252 Edge extraction unit 253 White region extraction unit 254 Color determination unit 255 Halftone dot separation unit 256 Total determination unit 261 First halftone peak detection unit 262 Second halftone peak detection unit 263 Halftone region detection unit (an example and color of halftone region detection means An example of using halftone dot area detection means)
H.264 halftone dot determination unit 271 character determination unit 272 expansion processing unit 273 decoding unit

Claims (12)

Image reading means for reading an image;
Chromatic area detecting means for detecting a chromatic area including colors other than black and white from the image data read by the reading means;
With reference to the two-dimensional small area in the image data read by the reading means, the absolute value of the density difference between the pixel of interest in the two-dimensional small area and the surrounding pixel group around the pixel of interest is predetermined. Comparison determination means for determining whether or not the threshold value is greater than
Halftone dot pixel detection means for setting the target pixel determined by the comparison determination means that the absolute value of the density difference is larger than the threshold value as a halftone dot peak pixel forming a part of a halftone dot;
Based on the relationship between the halftone dot peak pixel detected by the halftone dot peak pixel detection unit and the halftone dot peak pixel detected around the halftone dot peak pixel, a two-dimensional small region having a predetermined size including the halftone peak pixel A halftone dot region detecting means for detecting a halftone dot region as a halftone dot region,
The image processing apparatus according to claim 1, wherein the halftone dot region detection unit changes a halftone dot region detection result in accordance with a detection result from the chromatic region detection unit.   Of the color components included in the color image data read by the reading means, the first image data sensitive to the black density and the second image having high sensitivity to the color component having low sensitivity in the first image data. 2. A color-use halftone area detecting means for detecting a halftone area from color image data read by said reading means using at least two pieces of image data. Image processing apparatus.   The image processing apparatus according to claim 2, wherein the first image data is G image data for each of RGB color components.   4. The image processing apparatus according to claim 2, wherein the second image data is B image data for each of RGB color components.   The halftone dot region detection means includes a halftone dot pixel detection result in the first image data and a halftone dot peak in the second image data for the region determined to be a chromatic region by the chromatic region detection portion. 5. The image processing apparatus according to claim 2, wherein the two-dimensional small region having the predetermined size is detected as a halftone dot region using a logical sum with a pixel detection result.   The halftone dot region detection means includes a halftone dot pixel detection result in the first image data and a halftone dot peak in the second image data for an area determined as an achromatic region by the chromatic region detection means. 6. The image processing apparatus according to claim 2, wherein the two-dimensional small region having the predetermined size is detected as a halftone dot region using a logical product with a pixel detection result. An image input process for receiving an image signal;
A chromatic region detection step of detecting a chromatic region including a color other than black and white from the image data input by the image input step;
With reference to the two-dimensional small region in the image data input by the image input step, the absolute value of the density difference between the pixel of interest in the two-dimensional small region and the surrounding pixel group around the pixel of interest is previously determined. A comparison and determination step of determining whether or not the threshold is greater than a predetermined threshold;
A halftone dot pixel detection step in which the pixel of interest determined by the comparison determination step when the absolute value of the density difference is larger than the threshold value is a halftone dot peak pixel that forms a part of a halftone dot;
Based on the relationship between the halftone dot peak pixel detected by the halftone dot peak pixel detection step and the halftone dot peak pixel detected around the halftone dot peak pixel, a two-dimensional small region having a predetermined size including the halftone peak pixel And a halftone dot region detecting step for detecting a halftone dot region,
The halftone dot region detecting step changes the halftone dot region detection result according to the detection result of the chromatic region detecting step.   Among the color components included in the color image data input in the image input step, the second image having high sensitivity to the first image data sensitive to black density and the color component having low sensitivity in the first image data. A color-use halftone dot region detecting step for detecting a halftone dot region from the color image data input by the image input step using at least two pieces of image data with image data. 8. The image processing method according to 7.   In the halftone dot region detecting step, a halftone dot pixel detection result in the first image data and a halftone dot peak in the second image data for the region determined as the chromatic region by the chromatic region detecting step. 9. The image processing method according to claim 8, wherein the two-dimensional small area having the predetermined size is detected as a halftone dot area using a logical sum with a pixel detection result.   In the halftone dot region detection step, a halftone dot peak pixel detection result in the first image data and a halftone dot peak in the second image data for the region determined as an achromatic region by the chromatic region detection step 10. The image processing method according to claim 8, wherein the two-dimensional small area having the predetermined size is detected as a halftone dot area using a logical product with a pixel detection result.   An image processing program that causes a computer to execute processing according to any one of claims 7 to 10.   12. A recording medium on which the image processing program according to claim 11 is recorded.

Images