JPH0340078A - Color discriminating device - Google Patents

Abstract

PURPOSE:To decrease the influence by a color misalignment, to prevent the deterioration of discrimination accuracy of a color, and to utilize color format information as character segmenting information by averaging a chromatic color at every window of plural picture elements. CONSTITUTION:An image area P is divided, for instance, into windows W00 - W45 such as 16 picture elements in the horizontal direction and 4 picture elements in the vertical direction, and the hue of a chromatic color part in each window is averaged. In order to average the hue in such a way, an HSV converting circuit 10 for obtaining H, S and V signals from R, G and B signals, a comparator 11 for comparing the S signal with a threshold TS for separating to a chromatic color or an achromatic color, and a hue averaging part 12 for averaging the hue of a chromatic color in the window are provided. In such a way, the hue is converted to an average value, and a color distribution becomes narrow, therefore, color coding by a threshold is facilitated.

Description

[Detailed description of the invention] [Table of contents] Overview Industrial field of application Prior art (Figures 5 and 6) Means for solving the problems to be solved by the invention (Figure 1) Working examples (Figures 2, 3, and 4) Effects of the invention [Summary] Relating to a color identification device, format separation information is obtained by adding color to achromatic colors and changing color to chromatic colors at edge portions. In order to prevent noise from entering the original image signal, the original image signal is divided into hue (H),
A color identification device equipped with an HSV conversion means for outputting saturation (S) and luminance (V), including a hue averaging means for averaging the hue within a window section consisting of a plurality of pixels of an original image. It is characterized by

[Industrial application field]

The present invention relates to a color identification device, and more particularly to a color identification device used to accurately separate characters and color formats used in color documents such as OCR slips.

[Conventional technology]

With the development of office automation in recent years, color documents have come to be handled as document information. For example, frames are printed in different colors depending on the type of slip.

By the way, in OCR for reading handwritten characters, formats such as character entry frames are printed in a special color called drop-out color, and when reading, the format is optically removed and only the characters are read and recognized. That's what I was doing. However, with this method, it is necessary to select a filter that matches the spectral characteristics of the drop-out color and the writing instrument such as a pencil or ballpoint pen, and the seal becomes invisible when a red color is used as the drop-out color. Sometimes. Also, format information cannot be utilized.

To solve this problem, documents such as slips are read as three-color image signals, and saturation; hue; brightness (hereinafter referred to as S; 5atu
ration+H; Hue, V; Value
After dividing into achromatic colors and chromatic colors based on a predetermined darkness value using saturation, the luminance of achromatic colors is used for character recognition, and chromatic colors are converted into hues using multiple darkness values with predetermined hue values. An HSV identification method has been proposed that outputs a binary image or a multi-valued image corresponding to the number of colors.

By the way, in the HSV coordinates, the background, characters, character entry frame area, etc. are shown in the HSVS mark area as shown in FIG. That is, as shown in Fig. 5 (A), since the background and characters are achromatic, they fall within the area of the inner cylinder in HSV coordinates, and the colored character entry frame is chromatic, so they fall within the outer hollow cylinder. into the realm of. Utilizing this property after the 11SV conversion, first, the darkness value is multiplied by S to separate the character background, which is achromatic, from the character entry frame, which is chromatic. Then, a threshold value is applied to the separated achromatic color by V to distinguish the characters from the background.

On the other hand, for chromatic colors, as shown in Figure 5 (B), a darkness value for separating the hues is determined in advance for typical colors used on the slip, and this darkness value is used to divide the number of colors into a predetermined number. Identifies it as a multivalued image.

For this reason, conventionally, as shown in FIG.
2 will be provided. The image reading unit 41 has lenses 1, L2, and L on the lens.
3, a red filter RF that passes red light, a green filter CF that passes green light, a blue filter BF that passes blue light, and a charge coupled device (CCD) CL that detects red light. CC to detect
DC2, a photoelectric element for each pixel such as CCDC5 that detects blue light, scalers A1 to A3, and AD converters ADCI to AD that convert analog signals to digital signals.
C3, etc., and the color identification section 42 includes an HSV conversion circuit 50, comparison circuits 51 to 53, and a multiplexer 54.
．． 55 etc. are provided.

In FIG. 6, the upper lenses 1 to L3 of the image reading unit 41,
Red filter RF, green filter CF, blue filter B
F, CCDC1 to CCDC5, and AD converter AD
1 to AD3 constitute a three-system optical signal system of R, G, and B, read the image signals of the three colors of the original 43 line by line, and convert the image signals in the lines into digital signals for each pixel. , signal line RN, G7! , Bl these R
,,G.

The S signal is transmitted to the HSV conversion circuit 50. In the HSV conversion circuit 50, the S
Convert to signal, S signal, ■ signal and output.

Next, the S signal is input to the comparison circuit 51 and compared with a predetermined threshold T. As shown in FIG. 5(A), this threshold value T is used to distinguish between an achromatic color and a chromatic color, and when it is smaller than this, it is used as a character or background, and when it is larger, it is used as a color format. That is, if S<Ts, the comparison circuit 51 considers the color to be achromatic and outputs, for example, "0", and if S≧T, it considers it to be a chromatic color and outputs, for example, "1".

On the other hand, the ■ signal and the S signal are input to comparison circuits 52 and 53, respectively. A threshold value Tv is applied to the comparison circuit 52 in order to distinguish between the brightness of characters and the background among the achromatic colors, and when the ■ signal is smaller than this, it is determined as a black part, that is, a character part. A darkness value TH is applied to the comparison circuit 53 in order to distinguish colors.

As shown in FIG. 5(B), the threshold TH is T1 for distinguishing between red and yellow, T2 for discriminating between yellow and green, and T3 for discriminating between yellow and green.
is for distinguishing between green and cyan, T4 is for distinguishing between cyan and blue, and T5 is for distinguishing between blue and magenta.

The darkness value below T1 is red, and the darkness value above T5 is magenta. Therefore, as can be inferred from Figure 5 (B), if the character entry frame is one color, that is, the color of the format is one color, there is no need to differentiate, but if you want to distinguish between two colors, for example, two types of formats, the darkness value The darkness value, value, and number are determined based on the number of colors used on the slip, etc., such as 2 darkness values for 1 or 3 colors, 5 darkness values for 6 colors, and so on.

Note that the comparison circuit 52 compares the input V signal with the darkness value Tv, and if v<'rv, it treats it as a character and outputs, for example, "1", and if T≧TV, it treats it as a background and outputs, for example, "0". .

Furthermore, the output signal of this comparison circuit 52 is
The achromatic portion is output as is by the multiplexer 54 controlled by the output of 1, that is, when s<'rs, but the portion of the chromatic character entry frame when S≧T is output as is.
The "0" being passed to multiplexer 53 is output and removed in the same way as if it were read using a drop-out color. Note that the output signal S4 of the multiplexer 54 is transmitted to a subsequent character recognition section (not shown).

The multiplexer 55, contrary to the multiplexer 54,
When “1” is transmitted from the comparison circuit 51, the comparison circuit 5
3, that is, the character entry frame information is output as is, and when "0" is transmitted from the comparator circuit 51, that is, S<
When the color is T, that is, an achromatic color, "0", which is transmitted to the multiplexer 55, is output.

[Problem to be solved by the invention]

When the original 43 is actually read by the image reading unit 41, a flatbed scanner is used, so when reading in color, there is usually a positional deviation of about 2 pixels or 2 lines between the three colors (for example, RGB). Live. This is because, for example, when the RGB gradations are equal, the color of the pixel is achromatic, but for example, at the edge of a character, the color balance is disrupted and the pixel becomes colored, turning into a chromatic color. On the other hand, for chromatic colors, the color changes at the edges but never becomes achromatic. Therefore, when applying a threshold value based on saturation using the method shown in Fig. 4 above, the edge of the character will be mixed in as noise in the format information separated as chromatic colors, which has the drawback that it is difficult to use it as is as character extraction information. there were. Furthermore, because of the color shift, it was difficult to extract formats with specific colors.

Therefore, an object of the present invention is to provide a color identification device that solves these problems.

[Means to solve the problem]

In order to achieve the above object, in the present invention, as shown in FIG.
A window such as a pixel is divided into 0° to W4°, and the hue of the chromatic color portion within each window is averaged. Note that FIG. 1(A) shows a state in which 80×20 pixels are divided into 30 windows of 6×5.

For such hue averaging, as shown in FIG. 1(B), an HSV conversion circuit 10 obtains H, S, and V signals from R, G, and B signals, and separates the S signal into chromatic color or achromatic color. A comparator 11 that compares the chromatic color with a threshold value T, and a hue averaging unit 12 that averages the hues of chromatic colors within the window are provided. This hue averaging section 12 includes an AND gate 13, an adder 14, a line memory 15, a register 16, a memory counter 17,
It includes a memory control section 18 and an averaging section 19.

Now, the image area P in FIG. 1(A) is moved from window Wo0 to window W. Scanning is performed in the S direction, that is, the main scanning direction. At this time, 4 lines are scanned per window, so within the window. ~W0. is scanned in scan lines 10-13.

First, the R, G, and B signals obtained by scanning with the scanning line go are input to the HSV conversion circuit 10 shown in FIG. Ru. The S signal is input to the comparator 11, and the threshold Ts
The comparator 11 outputs "1" only for the larger pixel portion, that is, the pixel determined to be a chromatic pixel. Therefore, this chromatic H signal, that is, the H value, is input from the anime game 113 to the adder 14. One input of the adder 14 is rOJ for the first line 20, and since the line memory 15 has a memory capacity for one line of pixels in the image area, the line memory 15 corresponds to the window.

A chromatic H value corresponding to a pixel of ~13os is entered. Note that when processing this line (in the 0 direction), the memory counter 17 counts the number of output "1"s of the comparator 11. At this time, the memory control unit 18 calculates the count value of the memory counter 17 for every 16 pixels of line lO. read and retain.

Next is window support. ~W0. Similar processing is performed in the line 11 direction, but at this time, the line memory 15 holds the H value for one line of pixels obtained by the processing in the line 11 direction. Therefore,
Regarding the processing in the direction of line 11, first see Figure 1 (A
) at pixel Eo.

The H value of pixel Eo. When "1" is output from the comparator 11, the adder 1 is output via the AND gate 13.
4. At this time, pixel E of line memory 15
o. The H value for pixel E is temporarily held in the register 16, and is transmitted to the adder 14, so that the adder 14 ends up in the pixel E. The added value of 0 and EIO is output and held in the line memory 15. Then, the H value of pixel Eo from the line memory 15 is temporarily held in the register 16, so
Pixel E in adder 14; , and the H value of Elf are added,
It is held in the line memory 15. In this way, addition processing with line β0 is performed by scanning in the direction of line 11. Then, by processing in the direction of line 12, the H value of pixel E2° is transmitted to the adder 14 via the AND gate 13, where it is then added to (Eoo+E+). When scanning is performed in the line 13 direction in this manner, the line memory 15 holds the results of the four additions in accordance with the scanning position correspondence of the pixels. At this time, the memory control unit 18 stores scanning lines I2. -n3, data corresponding to the number of chromatic colors is held for each block. Therefore, the memory control unit 18 outputs data corresponding to the H value and the number of chromatic colors held in the line memory 15 for each window to the averaging unit 119.
Based on these, the averaging unit 19 calculates the average value of the H values for each window, that is, the average value of the hue, as will be described later.

[Effect]

According to the present invention, a hue average value is calculated for each window,
The chromatic parts within the same window are replaced with this average value. As a result, the hue is averaged, that is, even if a dark color exists discretely in a specific edge portion, it is diluted, that is, the color distribution is narrowed, so that it becomes easy to classify colors by darkness value.

〔Example〕

An embodiment of the present invention will be described based on FIGS. 2 and 3.

FIG. 2 is a block diagram of an embodiment of the present invention, FIG. 3(A) is a detailed diagram of a memory control section and an averaging section, and FIG. 3(B) 2 is a detailed diagram of a table.

In FIG. 2, the same symbols as those in FIG. 1 indicate the same parts, 1 is an image reading section corresponding to the image reading section 41 in FIG. 20 is a comparator which compares the V signal indicating the brightness with a threshold value T7, and if it is lower than this, it is regarded as a character part and outputs "1", 21 is an ant game 1
・When the comparator 20 outputs "1" and the comparator 11 determines "0", that is, an achromatic color, the inverter 22 outputs "1", thereby outputting "1". In other words, the AND gate 21 outputs a character signal "1".

Reference numeral 23 denotes a histogram memory, which stores the values averaged by the average value section 19 for the window size (Fig. 1 (A)).
In the example shown in , the histogram memory 23 is 16×4).
24 is a threshold value setting unit, which sets a darkness value in the valley of the hue histogram based on the histogram stored in the histogram memory 23, and creates a hue histogram. 25 is a lookup table, and the color (color number) is determined by referring to this lookup table 25 from the H value passed through the averaging unit 19 from the second document. It is. Note that the histogram setting and look assortment table creation procedures have already been patented by the present applicant, and are not directly related to color and stain correction, which is the subject of the present invention, so a detailed explanation will be omitted.

As described above, the present invention calculates the average value of the chromatic color portion of each window, and for this purpose, the number of pixels in the chromatic color portion of each window and the sum of the H value of this pixel portion are calculated. It is necessary to find out.

In order to obtain the number of pixels in the chromatic color portion within the window, for example, as shown in FIG. 31, a second counter 32 that counts the number of pixels 16 (0 to 15), which is the width of the window, and a table 33.
A control section 34 is provided. Of course, this first counter 31, the second
The counter 32 performs a counting operation in synchronization with the scanning of the line memory 15.

The memory counter 17 receives chromatic pixels from the comparator 11.
1", the first counter 31 is counting O when the line e6 is first scanned. The second counter 32 starts from O, and when it counts 15, the control unit 34 reads the memory counter 17, writes the count value in the table 33, and resets the memory counter 17. Therefore, Fig. 1 (A
), the chromatic color number Ao counted by the memory counter 17 for the first 16 pixels is entered in the count section Co, as shown in FIG. Then, the chromatic color number Bo counted by the memory counter 17 for the next 16 pixels is written into the second count section CI. In this way, when the number of windows in the horizontal direction for the scanning window is N, N pieces of data (Ao=No) for the scanning line 2o are written in the count sections C0-CN-1. Similarly, for the scanning line *t-7i3, each chromatic color number count value is entered in each count section, and data as shown in FIG. 3(B) is obtained. Note that writing data to this table 33 is as follows:
The first counter 31 gives β0 to β3, and the second counter 32 gives Go''-CN-+, thereby specifying the address.

When one row of windows in the horizontal direction has been scanned a predetermined number of times (four times in this embodiment) in this way, the control unit 34 scans the window W this time. . In order to obtain the number of chromatic colors in the table 33, data Ao, A+,
Ax and A3 are read and sequentially sent to the second calculation section 36 of the averaging section 19. As a result, the second arithmetic unit 36 calculates Ao+A1+At+A3 and holds this for the time being.

At this time, the memory control unit 18 sequentially outputs 16 pixel data corresponding to the window W0° from the line memory 15 to the first calculation unit 35 via the register 16.

This is performed under the control of the control section 34 in the memory control section 18. As explained with reference to FIG. 1, at this time, in each section of the line memory 15, the added values of the H values of the vertical pixels by the scanning line memory 15 are respectively written, so these are added by the first calculation unit 35. By doing so,
The sum of H values in the window W0° is obtained. Then, the sum of the H values is sent to the second calculation unit 36 and divided by the above-mentioned (A I + Aa + A3 + A, 4) to calculate the average value of the H values of the chromatic pixels that are equal to or higher than the threshold T6 in this window W0°. obtain.

Next, the control unit 34 selects the window W from the data memory 15 of the count classification CI from the table 33. 16 corresponding to 1
The data of each H value of the pixel is similarly sent to the first calculation section 35 and added together to obtain the sum of the chromatic pixels in the window W01, which is then added to the second calculation section 36 (Bo
+B++Bs+B3) to find the average value of the H values of the chromatic pixels that are equal to or higher than the threshold T5 in the window W01.

In this way, the window W in the horizontal direction. [1, After calculating the average value for WO+, the next column window w
Scan line 7 for oos w,,,W,2-!
Scan data in directions o to 13 are sequentially output. Similar processing is then performed on these, and the average value of the H values is determined for each window.

Note that when the added values for four lines are stored in the line memory 15, input of a new H signal is interrupted. Then, as mentioned above, the average value for the windows W0°, Wol, etc. is calculated, but when reading the first training manuscript,
The averaged value is divided by the window size (in this example, 16
X4) is output to the histogram memory 23 of the histogram creation section. When the averaging of the first four scanning lines is completed in this way, the next four lines are similarly averaged, and thereby the averaging process of the number of windows IWo, W1□- is performed. When the averaging of the entire image is completed, a hue histogram is created in the histogram memory 23. The darkness value setting unit 24 sets a darkness value in the valley of the hue histogram based on the histogram memory 23, and creates a look amplifier table 25. After passing through the averaging section 19, the color of the second document is determined by referring to the lookup table 25, but this is not related to the purpose of the present invention, so this histogram creation is not necessary. A detailed explanation of the look amplifier table will be omitted.

In the present invention, the means for calculating the number of chromatic pixels in the window is not limited to the means shown in FIG. 2 and FIG. .

That is, the memory counter 37 counts the number of chromatic colors within the window for each scanning line. At this time, the memory control unit 38 outputs "1" to the AND gate 39 for every 16 pixels to turn it on, so the AND gate 39 outputs "1" to the AND gate 39 for scanning line fio, for example, Ao in FIG. 3(B), Bo-・-N.

n is output and sent to the second line memory 4 via the adder 40.
1 is filled in sequentially. Note that the second line memory 41 has divisions in the horizontal direction according to the number of windows.

In other words, in FIG. 3(B), there are N divisions of Co-C, . Next, when scanning is performed by the scanning line 11, data Ao of the first section Co of the second line memory 41 is temporarily set in the register 42, and AI counted by the memory counter 37 is set to the AND gate 39.
The signal is input to the adder 40 via the . As a result, the adder 40 calculates (Ao+Ar) between Ao stored in the register 42 and this A+, and stores this in the section Co of the second line memory 41. When such control is performed on the scanning line *o-A3, (Ao+Al+A2+A9) is written in the section CO of the second line memory 4, and (Be + BI + BIl + 83) is written in the section C1. And in the category CN-1 (No+N++N2+
N3) will be entered. Therefore, the scanning line R
,NI! 3, the number of chromatic colors for each window is set in the second line memory 41 in this way, so the second calculation unit 36 of the averaging unit 19 simply performs the division. You can find the average value just by doing this.

Furthermore, when considering hardware implementation, it is desirable to be able to process with as little memory as possible.

Therefore, all processing such as HSV color conversion and averaging processing is performed line by line.

Although the window size has been described as an example of 16×4 pixels, it is of course not limited to this.

〔Effect of the invention〕

When reading a slip in color and separating black text from color format information, in the present invention, since there are only a small number of color misaligned pixels at the edge, the chromatic colors are averaged for each window of multiple pixels. , the influence of color shift can be reduced, color identification accuracy can be prevented from deteriorating, and color format information can be used as character segmentation information.

In addition, if color identification of the format is not required, saturation S and brightness V
Since it is possible to separate characters and formats by simply using the following commands, we can expect a significant improvement in processing speed.

[Brief explanation of drawings]

FIG. 1 is a diagram explaining the principle of the present invention, FIG. 2 is a configuration diagram of an embodiment of the present invention. FIG. 3 is an explanatory diagram of the main parts of the present invention, FIG. 4 shows another embodiment of the present invention, Figure 5 is an explanatory diagram of the HSV display system. FIG. 6 shows a conventional example. 0-HS V conversion circuit 1-Comparator 2-Hue averaging section 3-and gate 4- Adder 5- Line memory 6- Register 7-Memory counter 3 18-Memory control unit 19-Averaging section

Claims (1)

[Claims] A color identification device comprising an HSV conversion means (10) for outputting an original image signal in terms of hue (H), saturation (S), and brightness (V), comprising: a plurality of pixels of the original image; A color identification device comprising: hue averaging means (12) for averaging hues within a window section.