JPH0195673A - Digital color picture processor - Google Patents

Abstract

PURPOSE:To reproduce a black character with high fidelity by analyzing the color of an original into an R, a G and a B, reading it, extracting a black element, and deciding a character area for extracted black data. CONSTITUTION:The blocking circuit of an algorithm to decide by a colored picture element density possesses an MTF correction 1, a comparator 2, a colored picture element density filter 3, and a comparator 4. The algorithm is a separated method by using the fact that the density of the colored picture element is low at the black/white boundary part (an edge part) of the character, etc., the color of the original is analyzed to the red, the green and the blue (the R, G and B), read out, the black element is extracted, and the character area is decided for the extracted black data. Thus, the black character can be correctly extracted in respect to the picture data from the scanner of low costs and slightly deteriorating performance.

Description

[Detailed Description of the Invention] [Technical Field] The present invention relates to a digital color image processing device, and more specifically, it is applied to a digital color copying machine, a color facsimile, etc. that separates the black character image part of a color picture/text mixed image. The present invention relates to a digital color image processing device that can perform digital color image processing.

[Prior art]

In originals that contain both pictures and text, the text is often black even if it is a full-color original.To reproduce black on a full-color copying machine, three colors of yellow, magenta, and cyan (Y, M, and C) are required to reproduce black. Layer the color materials. However, even if the three colors Y, M, and C are layered, if the Y, M, and C colors are not perfectly balanced, some color components will remain. Furthermore, if the alignment of the Y, M, and C versions is incomplete, that is, if there is a positional shift, the image quality of images that require high resolution, such as characters, will be significantly degraded.

As a countermeasure for this problem, digital color copiers use
It has been considered to apply UCR (undercolor removal) in which the overlapping portions of the three colors , M, and C are replaced with M (Bk). However, in reality, the input system (scanner) is red, green, blue (R, G,
Even if 100% UCR is performed, it will not be converted to solid black due to the positional deviation between B) and differences in γ characteristics and MTF characteristics. Therefore, in order to reproduce black characters only in black, a method for determining the black character area was developed. It became necessary.

Conventionally, the determination of whether it is a black and white area uses data separated into R2O,B, and when R-G=B, black-and-white (
It was generally assumed that the color was achromatic. but,
In this method, as mentioned above, if a scanner with insufficient characteristics is used, the difference between R, G, and B (max.
(IR-G1.IC-B1.IB-R1)) may have a large value even for an achromatic chart, and it is difficult to accurately determine whether a pixel is a black and white pixel. On the other hand, improving the performance of Kisuyana would lead to higher costs and larger size of the device, which is not practical.

[Moonlike]

SUMMARY OF THE INVENTION The present invention has been made to improve the above-mentioned disadvantages of the conventional apparatus, and its purpose is to accurately print black characters even on image data from low-cost scanners with poor performance. An object of the present invention is to provide a digital color image processing device capable of performing extraction.

〔composition〕

In order to achieve the above-mentioned object, the present invention includes a reading device that separates and reads a document into red, green, and blue colors, and an extraction device that extracts a black component, and determines a character area based on the extracted black data. It is characterized by the following.

Hereinafter, 1. of the present invention. Examples will be described in detail based on the drawings.

First, the algorithm for separating pictures and text will be explained.

First, let's talk about the determination based on colored pixel density. In this determination, (1) When binarized at a low level (background level + α), pixels exceeding the background level - colored images can be extracted. .

(2) In general, it is thought that most areas of photographs (density level 1 originals) are composed of pixels at or above the background level. (
3) Since the character image is a black-and-white binary image, it is thought that the same amount of background-level pixels and high-density colored pixels coexist. T4) When the density of the colored pixels extracted in (1) is calculated, based on the assumptions (2) and (3), it is considered that photographic images have almost the highest level of density, and character images have a medium density. . Therefore, a threshold value is set, and when it is equal to or greater than the threshold value, it is determined to be a photographic image, and when it is smaller than that threshold value, it is determined to be a text image.

FIG. 1 is a block circuit diagram of an algorithm for determination based on colored pixel density. In FIG. 0, 1 is an MTF correction, 2 is a comparator, 3 is a colored pixel density filter, and 4 is a comparator.

Since this algorithm (1) is a separation method that uses the low density of colored pixels at black and white boundaries (edges) of characters, etc., erroneous determination is likely to occur if the edges are heavily rounded. Therefore, before extracting colored pixels, the MTF of the input system is
It is preferable to perform correction 1. The comparator 2 extracts colored pixels, and when the value is greater than or equal to the set threshold value thr 1, it is used as a colored pixel, and when it is smaller, it is used as a background pixel. Further, the coloring pixel density filter 3 in the next stage may have a size of about 4×4 to 8×8. As mentioned above, in a photographic area, almost all pixels in the scanning window are colored pixels, so the threshold value for area determination is approximately N-N-2 out of the N reference pixels in the scanning window. Just set it. The comparator 4 performs region determination, and as described above, the set threshold value t
If it is greater than or equal to hr2, it is a photo area, and if it is smaller, it is a text page area. In addition, if the scanning window size is 7 x 7, the reference pixels may not be all 49 pixels (as shown in Figure 2, (e), some of them may be used, such as a cross shape or an x shape). This also makes it possible to simplify the hardware.

Next, regarding the determination based on edge pixel density, (
1) Extract edge pixels using Laplacian operator, Roberts operator, etc., (2) Photographic images are 1 degree deep! In J, there are few pixels that are edge-extracted because there are few sudden changes in density. (3) Character images are basically binary images, so there are many pixels that are edge-extracted. (4) The density of edge pixels is When calculated, +21. [From 31, the density is lower in the photo area and higher in the text area.

FIG. 3 is a block circuit diagram of an algorithm for determination based on edge pixel density. In FIG. 0, 5 is a differential value calculation, 6 is a comparator, 7 is an edge pixel density filter, and 8 is a comparator. The comparator 6 performs edge extraction, and when it is greater than or equal to a set threshold value thr3, it is an edge, and when it is smaller, it is a non-edge.

The comparator 8 performs area determination and uses the set threshold value thr
If it is greater than or equal to 4, it is a text area, and if it is smaller, it is a photo area. The edge extraction method is Laplacian (2
A common method is to use the magnitude of the gradient (first-order differential) or gradient (first-order differential) value.

Figure 4 shows an example of a differential filter.
dl is a second-order differential filter, and (0) to (n) are first-order differential filters.

In edge extraction, the magnitude of the absolute value of the differential value is a problem, but since the first-order differential filters shown in (e) to (n) have strong directionality, (R) - (f), ( Illusion (h)
, (1)-(J), (g)! -(1), (IIri)
It is preferable to use the orthogonal differential value f of the directionality as shown in -(n), the square root of the sum of squares of Fx, and it is simply (lfxl+I fy I), /2
．． max (I fx 1.I fy l) may be used, or the maximum value of the gradients in eight directions at the pixel position of interest may be used.

In Fig. 5, if the gradation level of each pixel in the 3 x 3 scanning window is a x l, the chronic gradient (first derivative) at the pixel of interest position W (order a?) can be obtained by It will be done.

The determination level for edge extraction is a value that should be determined based on the MTF characteristics of the input system and the characteristics of the image to be extracted. The threshold value may be a fixed value, but a floating threshold value that takes into account the background level of the document may be used.The edge pixel density filter 7 is used to determine the M range of photographic areas and text areas. An advantage of using the edge pixel density is that it can prevent omissions in extraction of character areas and erroneous determinations in photographic areas due to the influence of noise. The size of the density filter is 3×3
It is sufficient to use a size of about 8×8.

Next, region determination based on halftone dot detection will be explained. This algorithm is only for extracting halftone photographic areas. (1) Slice and binarize the input image. (2)2
Detect halftone dots on the value image using a pattern matching method. (3) For each unit block, determine whether it is a halftone dot area based on the presence or absence of halftone dots. (4) The judgment is corrected based on the judgment results of adjacent blocks to reduce erroneous judgments due to noise.

FIG. 6 is a block circuit diagram of an algorithm for region determination based on halftone detection. In FIG. The comparator 10 has a set threshold (lII
It is binarized by thr 5, and the output of the judgment correction 13 is a halftone dot area signal. The halftone dot image is basically a binary image, and the dot diameter is modulated according to the size of 24 degrees to be expressed. This algorithm detects this dot and determines the area near the area where the dot is detected to be the halftone image area. A pattern matching method is used to detect dots, but since the dot diameter changes depending on the halftone dot pitch and the halftone dot m ratio, a plurality of types of templates are prepared.

′! In FIG. 57 showing an example of a template, each element within the scanning window is represented by MiJ, and three types of templates are represented by O9× and Δ. In this case, the halftone dot detection conditions are as follows.

Next, the image is divided into blocks each having a size of 8x8 to 16xl, and the block in which halftone dots are detected is determined to be a halftone dot area. In order to prevent misjudgments due to noise etc., the judgment is then corrected.As shown in Fig. 8, when three or more blocks out of four adjacent blocks are judged to be halftone areas, the remaining 1 block is also corrected as a halftone area. Conversely, if the number of blocks determined to be in the halftone dot jl region among four blocks is two or less, it is corrected that all four blocks are not in the halftone dot region. Again, each component of the apparatus for implementing the algorithm (1) for determination based on colored pixel density shown in the block circuit diagram in FIG. 1 will be explained in order. The image data read by a scanner (not shown) is
It is necessary to correct the deterioration of TF. Therefore, MTF correction 1 can be performed using a digital filter having high frequency emphasis characteristics.

The high-frequency strong g is usually calculated from the original image using the Laplacian (second-order differential).
A common method is to subtract. Figure 9 (
a), (b), and (C) each show examples of two-dimensional high-frequency strong filters used for MTF correction.The coefficients of these filters are actually determined according to the MTF characteristics of the input system. This is the value that should be used. Although FIG. 9 shows an example of a 3×3 size, more appropriate correction can be performed if a larger size filter such as 5×31 or 5×5 is used. Further, the coefficients in the main scanning direction and the sub-scanning direction may be configured to be different depending on the MTF characteristics of the actual input system.

Further, as shown in FIGS. 9(b) and (0), by using only pixels in four directions, hardware can be simplified.

FIG. 1.theta. shows a circuit diagram of a 3.times.3 filter. This circuit has a buffer for two lines and 3.times.3-9 latches in order to perform a 3.times.3 matrix operation. 1st
In figure 0,! 4.1.5 is line buffer, 16~
24 is a latch, 25 to 32 are adders, and 33 is a multiplier. The latched nine-pixel data is subjected to calculations as shown in FIG. 9 using adders 25 to 32 and a multiplier 33. Adders 25 to 32, instead of multiplier 33]'
? By using an OM (read-only memory), it is possible to easily perform matrix calculations on arbitrary coefficients.

Figure 1F shows the coloring density calculation circuit in Figure 0.
34 is a shift register, 3S is a counter, 36-39 are line buffers, 40-44 are latches, 45-48 are adders, and 49 is a comparator.

In the example shown here, the number of colored pixels within a 5×5 scanning window is counted. The low level 2 sulfurized 1-bit data is input to the shift register 34.

This shift register 34 outputs binary data for 5i elements (5 bits in total) in the main scanning direction.

The next stage counter 35 counts the number of colored pixels per five pixels in the main scanning direction. That is, the number of 1's in 5 bits is counted. The counter 35 can be realized using a ROM and a table reference method. It is sufficient to store a value from θ to 5 corresponding to the number of 1s in the 5 bits at the address addressed by the 5-bit data.Pixel data in the main scanning direction is held for 5 lines; The number of colored pixels per 5×5-25 pixels is calculated by adding the pixel density data for 5×5−25 pixels using adders 45 to 48. Comparing this colored pixel density with a predetermined threshold value thr (about 25 to 23),
It outputs an area determination signal of 0 (photo area) when it is larger than the threshold and 1 (text area) when it is smaller.

Algorithm for determination based on edge pixel density in Figure 3 (2
) block circuit diagram will be explained. Differential value calculation 5 is shown in FIG. 10, which determines whether or not it is an edge pixel by comparing the differential value output from the first stage differential value calculation circuit 5 with a predetermined threshold value (thr3). MT shown
This can be realized by using a circuit similar to the F correction circuit. The filter coefficients shown in FIG. 4 may be used at that time.

A circuit similar to the edge pixel density calculation and determination circuit can be used. At this time, if the edge pixel density is greater than a predetermined threshold (thr 4), 1 (character area)
And when it is small, O (picture area) is output.

Next, the block circuit diagram of the algorithm for region determination based on halftone dot detection shown in FIG. 6 will be further explained.

This algorithm (3) detects halftone dots from their shapes. Since the detection performance greatly depends on the MTF characteristics of the input system, it is desirable to perform MTF correction 9. Image data that has been subjected to MTF correction has a normal binarization level (approximately 32 for 6-bit data) thr
5, and detection is performed using a template as shown in FIG.

Figures 12(a) to 12(e) show examples of halftone dot detection circuits. Figure 12(a) is a circuit for aligning data of 49 pixels in a 7x7 scanning window, and is a line memory for 6 lines. 50-55
, 7x7-49 bits of latches 56 to 62. Binary data M, ~M1. for 49 pixels. can be taken out from the latch group 63 at the same time. 12th
(b), (C), (dl is a halftone dot detection circuit using pattern matching. It judges whether it matches the three templates 1 shown in Fig. 7, and the judgment result is no.
Outputs nX and nΔ.

n? outputs 01 when matching - l when not defeated @
Using the signals no, nX, and nΔ, a halftone detection signal n is obtained using the circuit shown in FIG. 12(e). n outputs 1 when a halftone dot is detected, and 0 when no halftone dot is detected.

FIG. 13 is an example of a region determination circuit.
is an 8-bit serial-parallel converter, 65 is the first OR gate, 66 is the first latch, 67 is the second OR gate, 6
8 is a second latch, 69 is a third latch, and 70 is a line memory. This circuit detects the presence or absence of halftone dots for each block of 8×8 size and outputs a halftone dot area signal P. 8×8
If at least one halftone dot is detected in the block, P-1
, 0, outputs po. yi point detection signal n is 8
It is input to a bit serial/parallel converter 64 and output every time eight pixels are accumulated in the main scanning direction.

Next stage first OR gate, 8 bits (g pixels) from 65
If at least one of them has l (halftone dot detection), 1 is set. That is, the presence or absence of halftone dots within the 8×1 block is determined. The output from this first OR gate passes through a first latch 66 and a second OR gate 67 to a line memory 70.
is memorized. This line memory 70 stores area determination results for every eight pixels of the first line. When the second line dot detection signal n is input, the first OR gate 65 outputs an area determination signal P for every eight pixels of the second line. At the same time, the line memory 70 outputs an area determination signal P for every 8 pixels of the first line, and the second OR gate 67
An 8×2 block area determination result is output. Similarly, the line memory 70 stores area determination results for 8×7 blocks. When the halftone detection signal n of the eighth line is input, the area determination result of the 8×1 block of the eighth line is inputted to the second OR gate 67 through the first latch 6G. The second OR gate 67 outputs a 3×8 block area determination result. 8x8 block judgment result 1
) is obtained, the third latch 69 holds the determination result. When the 8×8 determination result P is held in the third latch 69, the outputs of the first latch 66 and the second latch 67 are cleared, and 0 is written in the line memory 70.
Prepare for area determination of blocks on the 9th to 15th lines.

FIG. 14 is an example of a circuit for correcting all the results of the region determination shown in FIG. 13. The area determination signal P output from the circuit of FIG. 13 is stored in the first line memory 71 every eight lines. The area signal P stored in the first line memory 71 is output for every eight pixels and is held in the next latch 73. The same operation is performed for eight lines from the first line memory 71.

That is, the same signal is output while processing an 8×8 block. When the eighth (eighth line) data is output from the first line memory 71, the data is stored in the second line memory 72.

At that time, data of the next block (9th line) is stored in the first line memory 71. The first and second line memories 71 and 72 respectively store area signals of adjacent blocks in the sub-scanning direction. Region determination signals are output from the first and second line memories 71 and 72 for every eight pixels, and are held in the next latches 73 and 74.

The outputs from these latches are further transmitted to the next stage, latch 75.
7G is latched every 8 pixels. 4 latches 73-7
6 holds area signals of four adjacent blocks, as shown in FIG.

In order to prevent false judgments due to noise, the following Nantes gate 7
7 to 80 and the AND gate 81 to correct the judgment.
When one block is determined to be a bright halftone area, the spout block is determined to be a bright halftone area, and when there are two or less blocks, it is determined to be a non-halftone area.

Here, if it is determined to be a halftone bright area, 0 is output, and if it is a non-halftone area, l is output.

Next, a method for separating black text areas using the results of the cypress/text area determination for each color of R, G, and B will be described.

Even in the case of a color document in which ``-m'' and ``Kan'' characters are mixed, the characters are considered to be mostly black characters on a white background. R, G for a manuscript with black letters on a white background.

When the B color is separated and read, it is considered that each color has almost the same data. If the above-described picture/character separation processing (determination based on colored pixel density or determination based on edge pixel density) is performed on image data of each of these colors, almost the same determination results will be obtained. Furthermore, if similar separation processing is performed on a document with colored text other than black, for example, red text on a white background, the R light will not be absorbed by the red color material, so the text will be removed from the R image. There are images in which no area is extracted, and at least one character area for characters of other colors is not extracted. Therefore, if character extraction is performed for each color of R, G, and B as a feature of a black character on a white background,
It is thought that the same area is extracted for all three colors. In this algorithm, (1) R.

Picture/text areas are determined for each color of G and B. (2) R
.

The area determined to be a character area for all three colors G and B is determined to be a black character area. Figure 15 shows these R, G, B
It is a block circuit diagram of an algorithm that also uses picture/text area determination results for each color, and 82 is a character area determination circuit.

Next, the black character area is separated using wax (R+G*
B) A method using black character determination using data will be explained.

For the luminance data R, G, B, tsax (R, G.

I3) = -5in (positive, 9 pixels), that is, the minimum value of the density data tile, 1 row, represents the black component of that pixel.

I for achromatic color? ? G''4B, and R?0 for black pixels. Therefore, its complement R is close to the highest value. For chromatic pixels other than black, at least one of R, G, and H has a high brightness value. Therefore, wax (R+G, I3) is O
It was mentioned earlier that extracting the black component ([7CR) for every 0 pixel, which has a value close to , cannot be successfully performed unless the scanner characteristics are sufficient. However, the scanner's R, G, D
If the rLT(2) positional deviation is about 1/2 dot, it is possible to extract a black component of at least about 50%. In addition, for strongly colored pixels on a white background, even if the position is shifted, R,
Since there is almost no absorption for at least one of the colors G and B, the black component will not be extracted even if it is erroneous.This algorithm uses the black component data of this image. When picture/text separation (determined by edge ViA elementary density) is applied to the black component data l1aX (R, c, I3),
Black characters can be extracted. In this determination based on edge pixel density, the black component is extracted because several pixels around the edges of the image are determined to be bone regions.

Even if a certain amount of omission occurs, it can be compensated for when determining black characters. FIG. 16 shows a block circuit of an algorithm for determining black characters using wax (R+G, B) data, in which 83 indicates wax (R+G,B) data.

G, B) Calculation circuit 84 is a character area determination circuit.

Next, a description will be given of color correction of the color material of the document in separating the black text area.

Normally, if you want to reproduce a full color image, select Y.

Three colors, M and C, or four colors including Bk are used.

This is the same for silver halide photographs and for general printed matter using process inks. Here, the ideal characteristics of the Y, M, and C coloring materials are B, respectively.

The goal is to absorb 100% of G and R lights, and reflect (or transmit) 100% of other colors. However, in reality, except for Y, M and C have characteristics that are far from this ideal.
Table 1 shows the density of common process inks for R, G, and B light ("Color Reproduction Theory J.
P 32, by Yule, translated by Mawatari and Kokushi, Printing Society Publishing Department, February 5, 1971).

As seen from Table 1 P, 1, Y is relatively close to ideal, but C
, M each have a corresponding amount of ? 4. It can be seen that the Y component is included. On the other hand, when expressing colored characters in printing, Y
, M, and C. For example,
The green characters are expressed by overlapping C and )' ink. As mentioned above, C ink contains a considerable amount of H1 component, so it is an ideal green color (reflects 100% of G light).
I can't google it, but visually it is between green boxes, and there is no practical problem. However, this green letter is R, G, r
When reading with color separation into 3, G is added outside of R and B sights.
These R and G also absorb light.

If picture/text separation processing is performed on B data using algorithm (A) that uses picture/text area determination results for each color of R, G, and B, not only R, B, but also G data will be separated from text areas. It may be judged. Therefore, in this algorithm, R,G.

When B data is used, green characters may also be determined to be black characters. For the same reason, blue (C+M) color characters are also determined to be black characters. Also, max (R,G.

B) Algorithm (B) for determining black characters using data also extracts a considerable amount of black component from green characters, and similarly to algorithm (A), green characters are also determined to be black characters. Ru. Spectroscopically, these colored characters certainly have a black component, but as a copy image, the green characters should be reproduced only with C+Y and the blue characters (C+M), and the black component should be added. As a result, the saturation decreases, and the image quality deteriorates.Furthermore, if black characters are reproduced only in Bk, there is a risk that green or blue characters will be converted to black characters.

R, G in algorithms (A) and (B).

When B data is used, the above-mentioned problems can be reduced to some extent by adjusting parameters in the pictorial/character separation process. Algorithm (C) for color correction of the color material of the original in black character area separation solves this problem more easily and effectively.

In the process ink shown in Table 1, the transfer rate of each C, M, and Y ink is 1c, and the temperature is positive for M, Y, and R, G, and B light, and the relationship between T and T is as follows.

However, the values of 'C, M, and Y are 1 for the solid area and 0 for the background area.

-...-Formula l Therefore, the density Y of the original Y, M, and C versions of the manuscript.

M and C are given by the following equations by calculating the inverse matrix of the matrix in equation 1.

−・・−・Formula 2 R, G, B read by scanner (R, σ.

B) By performing the correction of Equation 2 on the data, color-separated image data for each of the Y, M, and C versions of the original can be obtained. Let's check this effect here.

The green characters on the document are expressed by superimposing Y and C solid images, so they are yc-t.

By substituting this into Equation 1, the values of the color separation data R, G, B (π, N) obtained from the scanner can be found.

It can be seen that there is considerable absorption of G light.

When this R, G, and B data is corrected using Equation 2, it can be seen that the green pixel is formed only from Y and C inks. In this example, the rows and columns of equations 1 and 2 are mutually inverse matrices, so this is a natural result. However, Y is commonly used.

The spectral characteristics of the M and C coloring materials are similar, and the correction value shown in Equation 2 can be determined using the average ink characteristic value. ) is different, but the difference is not large and it is sufficiently practical to use the average correction value. Furthermore, considering that the characteristics of the -m process ink and the toner and ink used in copying machines are not much different, the Y
, M, and C data can also be substituted. This eliminates the need to provide a separate color correction circuit, simplifying the hardware configuration. In addition, although a first-order color correction formula was used in this example,
A correction formula that uses even higher-order terms such as second-order and third-order terms may be used, or a correction formula for hue division that uses different correction values for each hue may be used.

Hereinafter, a copying system to which a digital color image processing device for separating black characters according to the present invention is applied will be described in FIGS. 17(a) and 1.
．． (This will be explained with reference to the block circuit diagram of C1, (d+).

In FIG. 17 (al is an example in which area determination using algorithm (A) is used), 85 is an input system (scanner), 86 is a shading correction circuit, 87 is an MTF correction circuit, and 88 is a γ correction circuit. , 89 is color/UCR
Processing circuit, 90 is wig (R, G, B-) calculation circuit, 9
1 is an area determination circuit, 92 is an intermediate processing circuit for the pattern section, 93
94 is a character processing circuit, 94 is a processing result selection circuit, 95 is an output control circuit, and 96 is an output system.The data read by the scanner (input system) 85 after being separated into R, G, and B is subjected to shading correction. , MTF correction, and γ correction.

The R, G, and B data that have undergone these corrections are color correction-U.
CR circuit 89, area determination circuit 91 and l1in (
R, G, B) 0 color correction input to calculation circuit 90 -
In the UCR circuit 89, color correction (masking) processing is performed according to the characteristics of the output system color material, and if necessary, OCR processing is also performed.
The M, C, and Bk data are input to a halftone processing circuit 92 for the picture area and a processing circuit 93 for the character area. The Bk data input to this character processing circuit 93 is min (R
.

G, B) Using data. This is because with UCR Bk data, even if UCR-100% processing does not necessarily extract black completely, this is to prevent black characters from becoming blurred or notches. R,
Either G or B may be used. The intermediate processing circuit 92 for the picture area may perform binarization processing (or multi-value processing such as 3-value or 4-value processing) that provides good grade characteristics, such as the systematic dither method or error diffusion method. The 0 character processing circuit 93, which carries out processing (hereinafter referred to as binarization processing as a representative example), performs processing with emphasis on resolution so as not to impair the sharpness of characters and line drawings. Processing methods that emphasize resolution include dither processing using a distributed pattern such as Bayer type and binarization processing using a fixed threshold. For these two types of binarized data, the processing result selection circuit 94 selects the processing result for the text part for the text area and the processing result for the picture part for the picture area, according to the result of area determination. In addition, in order to reproduce black characters with only black, in the area determined to be a black character by the output side circuit 95 in the next stage, Y, M
.

Control not to output C. The area determination circuit 91 first uses algorithm (1) or/and algorithm (2) and algorithm (1) for each R, G, and B data.
In (2), the area is determined to be a picture area, or the area determined to be a halftone dot area by algorithm (3) is determined to be a picture area. The remaining area is determined to be a character area. Regarding black, the area determined to be a character area for all 83 colors, R, G, is defined as a character area. The remaining area is determined to be a picture area.

FIG. 17(b) is a circuit diagram when black character area determination using algorithm B is used. The difference from FIG. 17(a) is that the area determination circuit 91 uses wax (R+G, B)
The point is that wax (R, G, B) data from the calculation circuit 97 is used. In this example, max (R, G, B
) is calculated, but R input to the area determination circuit 9I
, G, B data to calculate max(R,G
, B) may be calculated. In this circuit, black s
+ax (R, G, B) data is used to determine picture/text areas for black data using algorithm B. For R, G, B data, algorithm (1) ~
Picture/text area determination is performed using the combination of (3). Further, based on the R, G, and B determination results, black character determination may be performed using algorithm (A). At this time, algorithm (A).

It is only necessary to perform black character judgment using the logical sum of the results of (B). For black characters on a white background, algorithm (A
) can more accurately judge the N range, but the algorithm (A
), a black character on a colored background is not determined to be a black character.For example, if there is a black character on a yellow background, the B data is a solid image and there is no character area. Therefore, R,
Even if G is determined to be a character area, since B is not a character area, it is not determined to be a black character area. algorithm(
In B), since black component data is used, even in the case of a black character on a colored background, there is no black component in the background (yellow), so the same result as in the case of a black character on a white background can be obtained regarding the black component data.

Therefore, in algorithm (B), even black characters on a colored background are determined to be black character areas. In this way, by using both algorithms (A) and (B), black character areas can be determined more accurately.

FIG. 17(C) shows R, G,
Not B, but Y whose color has been corrected taking into account the spectral characteristics of the ink.
, M, C data, which is different from the circuit diagram shown in FIG.
, B) is a calculation circuit. In this example, Y, M, and C colors are corrected for the output system 96 to simplify the hardware.
I'm trying to use data instead.

18(a) and crest) respectively show block diagrams of the primary color correction processing circuit.
~99 are latches, and 100~108 are ROMs.

109 to 111 are adder circuits, and 112 to 114 are latches. The first-order color correction equation is expressed as Equation 5.

From Equation 5, the C, M, and Y values after correction are as follows.

Cma,, *π+a r+ *7C+ a 3. *100・-・・・・Formula 6-(11Mm, a(z*-R+a
zt*G+ast*100...-Formula 6-(2) Y=
a, z*R+a,, *G+a,, *B・-・-=Formula 6-
(3) In the example of Fig. 18(a), Equation 6- (1) ~ (3
The 9th term on the right side of 1 can be multiplied by a table lookup method using ROM. The result is stored, and the value of aB*G can be obtained at the address addressed by G.

Addition circuits 109 to 111 use equations 6-(1) to 6, respectively.
- Add the three terms on the right side of (3). As a result, C, M, and Y values are output from adder circuits 109 to 111, respectively.

18), 115 to 117 are latches, 118
~120 is ROM, 121 is adder circuit, 122~127
is a latch, and 128, 129, and 130 are delay devices. In this circuit example, the calculations of equations 6-(1) to 6-(3) are performed in a time-division manner, thereby reducing the number of ROMs and adder circuits. In this circuit, one pixel clock is divided into three,
Every 1/37 clock, the formula 6-(13, 6-(2),
6-(3) is performed.

The operation will be explained with reference to the time chart shown in FIG. 18 (0). R, 'G, and "T data before zero color correction are latched to latches 115 to 117, respectively, in synchronization with the pixel clock.
It can be latched to. The ROMs 118 to 120 have 0 color selection symbols S1 . So is 1/
ROM1 repeats 00, Of, 10 every 3 clocks
At 18, the color selection signal is 00. Of, corresponding to 10, a,
, *R* a lt ” Rr 8131' Outputs the value of R. Similarly, in ROM119.120, a□*
Finished, a0*te.

Output the values of ats*G, asr*B, a3z*B, a;, *1, respectively. From the adder circuit 121, the ROM
The sum of the output values of ROM 118 to ROM 120 is output.

That is, color correction data C, M, and Y are output in correspondence with color selection signals of 00.01 and 10. Addition port 61.21
The C, M, and Y values are output in a time-divided manner. latch 1
22 to 124 each latch the C, M, and Y values output from the adder circuit 121 in synchronization with clocks whose phases are shifted by 1/3 clock. Latch 125-12
7, the C, M, and Y data are latched at the timing when the C, M, and Y data of the same phase are output from the latches 122 to 124. As a result, C and M with the same phase (same pixel)
, Y data can be referenced simultaneously from latches 125-127.

Figure 19 shows that in algorithm (B), the black component akin (Y, M,
The system is configured to use data extracted from C).

The difference in the circuit diagram from FIG. 17(a) is the color correction circuit 8.
9' and lIi n (Y e M + C) calculation circuit 97'. As in the case of FIG. 17(C),
The configuration may be such that both algorithms (A) and (B) are used.Also, if the black component extraction data is subjected to 100% UCR in the UCR circuit 89, the UCR
Bk data output from the CR circuit 89 may also be used.

When extracting a black character area using algorithm (B), black characters on a colored background are also extracted. At this time, 1m1n
An image based on (R, G, B) data becomes a solid image.

Therefore, when extracting black characters using algorithm (B), the character processing is a+ax (R).

G, B) or m t n (Y + M + C
) must be used. Also, algorithm (A),
(B) If both are used, the area determined as a black text area only by algorithm (B) can be determined as a black text area on a color background, so in this case, max (R, G.

Using algorithm (B), the area determined to be black by algorithm (A) (black characters on a white background) is win (R,G.

If B) is used, a good black character image can be obtained.

While referring to FIGS. 20(a) to (d), FIG.
al~(C1 and the processing switching operation of the circuit in FIG. 19 will be explained.
The same parts as in FIGS. (a) to (C) and FIG. 19 are given the same reference numerals.

In the example shown in FIG. 20 [a], the picture/character processing is switched only for Bk. The black character area signal is obtained by algorithm (A) or (B).

In algorithm (A), the black data I3k' input to the character processing circuit is ll1in CR,
G.

B), eaax (Y, M, C), R, G, B, etc. can be used. For the case of algorithm (B), max (R, G, B) or m
It is preferable to use in (Y, M, C),
Furthermore, the output control unit 95 is configured not to output YMC in the black character area, so that black characters are reproduced in only black. FIGS. 21(a) and 21-) show black character area signal generation circuits based on algorithms (A) and (B), respectively. In these figures, each color's character area signal B
or Y (LR3), G or M (LR3) and R or C (LR3) are algorithm (1) or/
And according to algorithm (2), halftone area signals B or Y (NR3), G or M (NRS) and R
Or C(NRS) can be obtained by algorithm (3).

In the example shown in FIG. 20(b), the picture/character processing is switched not only for black but also for Y, M, and C. Other configurations and operations are the same as those in FIG. 20(a). Figure 21 (character area signals of each color in el (LR3
) and an example of a black character area signal generation circuit based on algorithm (A).

In the example shown in FIG. 20 (Ct), Y, M, and C are not outputted only in the case of black characters on a white background (referred to as white and black characters). The algorithm (
When obtained by B), black characters on a colored background are also extracted.

When algorithm (B) is used in the example shown in FIG. 21, an output image in which the periphery of the black character is white is obtained for a black character on a colored background.

In the example shown in FIG. 20(C), the background color is prevented from appearing white against black characters on a colored background. Black and white character areas can be obtained by algorithm (A).Black characters can be obtained by algorithm (B), regardless of whether it is on a zero-color background or a white background, but in the example of FIG. I try to use logical sum with the result. In the example of Fig. 20 fcl, if we consider black characters on a color background as data for Bk literature processing, tax (
R, G, B) or n+in (Y, M, C
) It is better to use data.

Figure 20 (In the example of dl, the output of Y, M, and C is only in the black and white character area, so that Y, M, and C are not output, and at the same time, for black and white characters, win (R, G, B) is set.
.

max (Y, M, C), R, G, B, Y, M, C (
Select the processing data for the character part using (Y, M, C are non-UCR data), etc., and use wax (R, G, B) or mln for characters on colored ground (other than black and white characters).
(Y+M*C) data is used. This allows wax (R+G, B
)% sin (Y6M, C), it is possible to eliminate the possibility of fading of black characters and the appearance of notches, etc., and also to reproduce black characters on colored backgrounds. The black and white character and black character area signals may be obtained from the circuit shown in FIG. 21(d). In FIGS. 20(a) to (d) and FIGS. Area signal, NR3
indicates a halftone dot area signal, the output of the text area signal is 1 for characters, 0 for a picture, and the halftone dot area signal is 1 for text (non-halftone area) and 0 for a photo (halftone area). ) is shown.

〔effect〕 According to the present invention, the original is R and G.

The system is equipped with a reading means that separates the colors into B and reads them, and an extraction means that extracts the black component, and determines the character area from the extracted black data. It is possible to provide a digital color image processing device that has the effect of being able to reproduce black characters with high quality by reproducing only black characters.

[Brief explanation of the drawing]

Figure 1 is a block circuit diagram of an algorithm for determination based on colored pixel density, Figures 2 (a) to (f) are schematic diagrams showing pixel patterns, and Figure 3 is a block circuit diagram of an algorithm for determination based on edge pixel density. , Fig. 4(a)-(n
) are schematic diagrams showing an example of the differential filter used in the circuit of Figure 3, Figure 5 is a schematic diagram showing an example of the size of the density filter of Figure 3, and Figure 6 is a schematic diagram showing an example of the size of the density filter in Figure 3. A block diagram of the determination algorithm, FIG. 7 is a schematic diagram showing an example of a template used in the circuit of FIG. 6, FIG. 8 is a schematic diagram illustrating correction of the halftone dot fiJj area, and FIG.
, (kl), and (C) are schematic diagrams showing examples of two-dimensional high-frequency emphasis filters used for 'MTF correction, and the first
Figure 0 is a block diagram of a 3x3 filter circuit, Figure 11 is a block diagram of a colored pixel density calculation circuit, and Figure 12 (81-
(14) is a schematic diagram showing an example of a halftone dot detection circuit, FIG. 13 is a schematic diagram showing an example of an area determination circuit, and FIG.
A schematic diagram showing an example of a circuit that corrects the results of the circuit shown in the figure. Figure 15 is a block diagram of an algorithm that uses the pictorial/text area determination results for each color of R, G, and B. Figure 16 is
Block circuit diagram of algorithm for black character determination using ax (R, G, B) data, Figure 17 (a) to (C)
) and FIG. 19 are block diagrams showing circuit diagrams of a copying system to which the digital image processing device for black character separation according to the present invention is applied, and FIGS. 18(a) and (b) are examples of primary color correction circuits, respectively. A block diagram showing Fig. 18 (
C) is a time chart showing the operations in FIGS. 18(a) and (b), and FIGS. 20(a) to (d) are time charts showing the operations in FIGS. 17(a) to (
C) and a circuit diagram for explaining the switching operation of each process of the circuit shown in FIG. 19, FIGS. 21(a) to 21.
(d) is a block diagram showing a black character or white character trM area signal generation circuit according to each algorithm. 3... Colored pixel density filter, 7... Partition pixel density filter, 11... Halftone detection circuit, 85... Input system, 89... Color correction (UCR) circuit, 90... Calculation circuit , 91... Area determination circuit, 94... Processing result selection circuit, 95... Output control circuit. Figure 1 Figure 2 (a) (b) (c) (d)
(e) (f) Figure 3 Figure 4 (a) (b) (c) (d) (e)
(f) ((]) (h) (i)
(j) (k) (1) (m) (n) Fifth
Figure 15 η Figure 16 (e) (C) (b) Work O's Q Σ Σ Cr: Φ01 Figure 20 (a) Figure 20 (b) Figure 20 (C) Rod ? 4

Claims (1)

[Claims] A digital color image comprising a reading means that separates and reads a document into red, green, and blue colors, and an extraction means that extracts a black component, and determines a character area from the extracted black data. Processing equipment.