JPH11110551A - Color-discriminating device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To optimally remove a noise (false color) in a black area caused by mis-discrimination of colors. SOLUTION: Colors of respective pixel positions of a colored image fetched by dividing into color components R, G, B by a CCD 23 are discriminated in a color discriminating part 255 and the noise due to in the black area is removed in a black area correcting part 256. Color-discriminating data is extracted by the specified number of pixels in the black area correction part 256, and the number of pixels with the same color is counted in a circuit 256a to count the number of pixels with the same color. Furthermore, a distribution degree for indicating the degree of distribution of the colors is calculated by using a counting result in a distribution degree calculating part 256b, and an extracted area is discriminated whether it is black or not by using the distribution degree in a black discriminating circuit 256c. Then when the extracted area is not black, the noise is corrected by replacing the color discrimination data of the pixel at the center of the extracted area into black data. The noise is removed properly by discriminating presence/absence of the noise by checking the distribution of the colors by a specified number of pixels.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color discriminating apparatus for discriminating a color of an image which has been separated into image data of color components such as red, green, blue and the like using the image data of the color components. Things.

[0002]

2. Description of the Related Art Hitherto, various methods for discriminating the color of an image formed of an electric signal have been proposed. For example, Japanese Patent Laid-Open No. 5-1
JP-A-37011 discloses a method of calculating the difference between the maximum value and the minimum value of the image data Xr, Xg, and Xb of each of the R, G, and B color components.
If the calculation result is smaller than a predetermined threshold value, it is determined that the color is achromatic.
A method of discriminating a chromatic color based on a combination of magnitudes of image data Xr, Xg, and Xb of respective color components of G and B is shown. If it is determined that the color is achromatic, the luminance data is further used to determine white and black. In addition, the relationship between the color discrimination result once and the color discrimination result of the surrounding pixels is checked, and erroneous discrimination of each pixel position is corrected by majority decision.

[0003]

In the case of performing color discrimination of a color image from image data, if a discrimination criterion is set with emphasis on color discrimination of a chromatic portion in consideration of color reproducibility, black and chromatic colors are determined. , And the black pixel position is likely to be erroneously determined to be a chromatic color. For example, when a character or figure drawn on a whiteboard is photographed as a recording material by a digital camera, R, R, and R are added to pixels in a black area depending on the lighting conditions of the Bodo, the degree of dirt on the pen, and the performance of the camera.
A deviation may occur in the image data of the G and B color components, and the pixel may be erroneously determined as a chromatic color. As a result, a problem occurs in which chromatic colors are mixed as noise in the black region of the captured image.

[0004] If chromatic colors are mixed in the black area of the photographed image,
Since characters and the like are difficult to see and the image quality is significantly reduced, it is desirable to avoid mixing chromatic colors as noise in the black area in the color discrimination processing as much as possible. However, if the criterion for discriminating black and chromatic colors is strict, conversely, achromatic colors will be mixed with noise as chromatic colors, so it is difficult to prevent the occurrence of noise in the black region only by color discrimination processing. is there.

[0005] Therefore, when performing color discrimination processing with emphasis on color discrimination of chromatic color portions, a process of removing chromatic colors contained in a black region from the color discrimination result (hereinafter, referred to as noise removal process). )Is required.

The noise removal processing in the above-described conventional color discrimination method performs noise removal by majority decision with the result of color discrimination of peripheral pixels. Therefore, it is effective when noise is scattered in a black area. However, when the color discrimination result of the black area is occupied by a large number of noises such as spots, pixels that are correctly determined to be black are converted to chromatic colors due to the influence of peripheral pixels that are erroneously determined to be chromatic. It is highly likely that it will be incorrectly corrected,
It is difficult to correct the black area to black. In particular, in an image obtained by photographing a white boat with the digital camera described above, the black area is likely to be erroneously determined to be a spotted pattern due to the influence of lighting, stains on a pen, and the like. However, it is difficult to obtain a sufficient effect even if the noise removal method is applied.

The present invention has been made in view of the above problems, and provides a color discriminating apparatus capable of effectively removing noise generated in a black area by color discrimination.

[0008]

SUMMARY OF THE INVENTION The present invention is a color discriminating apparatus for discriminating the color of a color image composed of image data of a plurality of color components. A color discriminating means for discriminating, and extracting data of the color discriminated by the color discriminating means in units of a predetermined number of pixels, and calculating a degree of dispersion indicating a degree of dispersion of the color included in the extracted area. Calculating means for calculating, a determining means for comparing the degree of dispersion calculated by the calculating means with a predetermined threshold value to determine whether or not the color variance is large; Color data replacement means for replacing the color data at the pixel position at the center of the extraction area with black data when it is determined to be large (claim 1).

[0009] According to the above configuration, in a color image, the color of each pixel position is determined using the image data of the color components constituting the image. Further, color data of a partial area of the color image is extracted from the color determination result in units of a predetermined number of pixels, and the degree of dispersion of the color included in the extracted area is calculated. Is compared with a predetermined threshold value set in advance to determine whether or not the color variance is large. When it is determined that the color variance is large, the color data at the pixel position at the center of the extraction area is replaced with black data.

The above-described data replacement processing is sequentially performed for each pixel while changing the extraction area in the color image, whereby the chromatic color occurring as noise in the black area is corrected to black.

Further, in the color discriminating apparatus according to the present invention, when the color data replacing means determines that the color variance is large by the discriminating means, the color data replacing means adds the color data at the pixel position at the center of the extraction area to the color data. Instead, the color data of the entire extraction area is replaced with black data.

According to the above configuration, when it is determined that the color variance is large, the data of all colors included in the extraction area is replaced with black data. This makes it possible to speed up the process of correcting the noise in the black area.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described by taking a digital camera provided with a color discriminating apparatus according to the present invention as an example.

FIG. 1 is a perspective view showing an external appearance of an embodiment of a digital camera provided with a color discriminating apparatus according to the present invention, and FIG. 2 is a rear view of the digital camera.

In FIG. 1, a digital camera 1 is provided with a photographing lens 2 comprising a zoom lens substantially at the center of the front surface, and a light projecting window 4 for measuring a subject distance by an active distance measuring method and a light receiving window above the photographing lens 2. A window 5 is provided, and a photometric window 3 for measuring the brightness of the subject is provided between the two windows. Further, a finder objective window 6 is provided on the left side of the light projecting window 4.

The light projecting window 4 is a window for irradiating the subject with infrared light, and the light receiving window 5 is a window for receiving the reflected light of the infrared light from the subject. In the present embodiment, the active distance measuring method is adopted as the distance measuring method, but a passive distance measuring method may be used.

On the side of the digital camera 1, there is provided a card insertion slot 7 into which a hard disk card 10 (hereinafter abbreviated as HD card 10) is inserted and ejected, and the HD card 10 is ejected above the card insertion slot 7. A card removal button 8 for performing the operation is provided. Further, a shutter button 9 is provided at the left end of the upper surface of the digital camera 1.

When the photographing result is printed out, the card removal button 8 is pressed to remove the HD card 10 from the digital camera 1, and the HD card 10 can be mounted on a printer to which the HD card 10 can be mounted and printed out. .

The digital camera 1 is provided with a SCSI cable interface, and the digital camera 1 and the printer are connected by a SCSI cable.
, The image data may be transferred to a printer and the photographed image may be printed out directly.

In this embodiment, a hard disk card conforming to PCMCIA is used as a recording medium for image data. However, if a photographing result can be stored as image data, a memory card or a mini disk (MD) can be used.
And other recording media.

As shown in FIG. 2, on the rear surface of the digital camera 1, a main switch 11 for turning on the power is provided at the upper left end, and a finder eyepiece window 12 is provided substantially at the center. A color correction switch 13, a noise correction switch 14, and a density setting switch 15 are provided below the main switch 11.

Further, at the right end of the back of the digital camera 1,
A mode change switch 16, a color registration confirmation display section 17, a color selection button 18, a set button 19, and a default button 20 are provided.

In the case where characters, diagrams, and the like written on a whiteboard are directly photographed as recording materials, the photographed image is information in which characters and the like are clearer than the image quality having high descriptiveness as in ordinary photographing. High quality image quality is required. The color correction switch 13 is used to output characters,
A switch for instructing predetermined image processing for clarifying the density of a diagram or the like and clarifying the color.

In this image processing, characters, diagrams, and the like are usually written on a whiteboard with several types of color pens, such as red, blue, green, orange, and black. Since the importance of the information (eg, legibility, legibility, etc.) of the characters in the captured image is more important than the color reproducibility, the characters, etc. in the captured image are red, blue, green, orange, black, etc. Of which color is colored, and assigning a predetermined color according to the determination result to the character or the like (the image data of the region constituting the character or the like is Image data).

When the color correction switch 13 is on, the above-described image processing is performed on the photographed image. When the color correction switch 13 is off, the above-described image processing is not performed on the photographed image, and image processing is performed on normal photographing. Therefore, if the shooting mode for performing normal photographing is “normal mode” and the shooting mode for recording and shooting characters and the like written on the whiteboard is “document mode”, the color correction switch 13
Is a switch for switching and setting between the normal mode and the document mode.

The noise correction switch 14 is a switch for correcting a false color generated as noise in characters or the like of a photographed image in the image processing in the document mode. When the noise correction switch 14 is turned on, a process of correcting a false color generated in a character or the like of a captured image is performed as described later, and when the noise correction switch 14 is turned off, the correction process is not performed. Since the noise correction switch 14 functions in the document mode, even if the noise correction switch 14 is turned on while the color correction switch 13 is turned off, the false color correction processing instruction is ignored.

The density setting switch 15 is used to set the density information of the subject in the image processing in the document mode so that, for example, characters with thin lines, figures with low density, etc. can be formed into clear images. The density information set by the density setting switch 15 is used when a determination threshold in color determination in image processing in the document mode is switched according to the density of the subject. In the present embodiment, the density setting switch 15 switches between “dark” and “light” and sets two types of density information. Since the density setting switch 15 also functions in the document mode, the color correction switch 1
Even if the density setting switch 15 is turned on while the button 3 is off, the instruction to correct the density of characters and the like is ignored. Note that the density setting may be set continuously.

The mode change switch switches between a photographing mode for photographing and a color registration mode for registering a discrimination color in color discrimination. In the color registration mode, a color sample (such as a color patch created on a whiteboard with a color pen) is actually photographed, and image data of colors used in the above-described color discrimination processing and color replacement processing is set and registered. This is a mode for performing processing. The digital camera 1
Is set in advance in the color registration mode. If the user does not register and set the color data in the color registration mode, the default value is used for color discrimination and color data replacement processing.

In the present embodiment, as shown in FIG.
It is possible to register four colors of red, blue, orange, and green. Therefore, the mode change switch 16
Is set to “Color Registration”, red, blue, orange, green
For each color, a color sample can be taken and the desired color data can be registered.
When set to, photography is possible.

The color registration confirmation display section 17 displays the contents of color registration and the colors used in the color data replacement processing. The color registration confirmation display section 17 serves as a display section for selecting a color to be registered in the color registration mode, and a display for selecting a color used for color data replacement processing in the shooting mode. Also functions as a unit. FIG.
, The indications of “red”, “blue”, “orange”, and “green” indicate the types of colors that can be registered. In the present embodiment, four types of colors can be registered. However, four or more types of colors can be registered, and colors other than the above four types of colors (for example, colors such as yellow and brown) can be used. You may make it possible to register.

A color with a bar display 171 at the top indicates that the color is selected when color registration or color use is selected. In FIG. 3, “blue” is selected. The bar display 171 is moved left and right by operating the color selection button 18, and the user can select a desired color by operating the color selection button 18 in the color registration mode. Each time the right color selection button 18 is pressed, the bar display 171 moves rightward as “red → blue → orange → green → red”, and each time the left color selection button 18 is pressed, the bar display 171 is displayed. `` Green ← red ← blue ←
Move to the left as “orange ← green”.

A color surrounded by a circle 171 in the color registration column indicates that the color has been registered, and a color not surrounded by a circle 171 indicates that the color has been registered. It is not shown. The color registration is performed by setting a color registration mode, selecting a desired color, for example, “red” on the bar display 171, and then filling an area of an appropriate size on the whiteboard with a red pen or the like. This is done by taking a sample. In this case, as shown in FIG. 4, a frame 21 indicating the color data capture range is displayed in the center of the viewfinder, and a frame is displayed so that the color sample 22 is included in a certain area or more in the frame 21. When the photographing is performed with the adjustment of, color data registration processing is performed. When shooting in color registration mode,
Using the image data included in the frame 21 in the captured image,
R, G, B constituting the color (red in the example of FIG. 4) in the frame
Is registered (stored in the memory) with the second-order normalized data (to be described later) generated by using these color components.

The color use column is a display indicating whether a registered color or a default color is used in the process of color determination and color data replacement in the color correction mode. ○ mark display 1
The color surrounded by “ON” in 73 indicates that the color is used, and the color surrounded by “OFF” indicates that the color is not used. The display example of FIG. 3 shows that three colors of red, blue, and green except for orange are used for the color data replacement processing.

The ON / OFF setting of the color use column is performed by operating the set button 20 after selecting a color with the color selection button 18 in the color registration mode. Each time the set button 20 is pressed, the position of the mark 173 in the color use column is switched between ON and OFF alternately. The user operates the set button 20 while watching the display position of the mark 173. Use / non-use of color can be set.

The default button 20 is an operation button for returning a registered color to the default color. In the color registration mode, when the desired color is set on the bar display 171 and the default button 20 is pressed, the display of the mark 172 disappears, and the color registration is deleted.

FIG. 5 is a block diagram showing the image processing of a photographed image in the document mode of the digital camera according to the present invention. In the figure, the same members as those shown in FIGS. 1 and 2 are denoted by the same reference numerals. The imaging unit 23 converts a subject light image formed on the imaging surface of the imaging device by the imaging lens 2 into an image signal and outputs the image signal. The image pickup unit 23 includes an image pickup device composed of, for example, a CCD color area sensor and this image pickup device (hereinafter, C
It is called CD. ), And an analog signal processing circuit that performs predetermined signal processing such as noise reduction and level adjustment of an image signal (analog signal) output from the CCD.

The CCD color area sensor separates the subject light image into image data of R, G, and B color components and outputs the image data. The CCD color area sensor may be a single-plate area sensor or a three-plate area sensor. . The imaging unit 23 controls the CCD imaging operation (charge accumulation and readout of accumulated charges) based on the exposure control value (shutter speed) input from the control unit 34 that centrally controls the imaging operation of the camera. That is,
Exposure (charge accumulation) of the CCD is started based on an exposure start signal from the control unit 34, and when the exposure is completed after a lapse of a predetermined time, the accumulated charge is transferred to the CCD for each of the R, G, and B color components.
After being subjected to predetermined signal processing by an analog signal processing circuit, the signal is read out to an image processing unit 26 via an A / D converter 24.

The A / D converter 24 converts an analog image signal read from the imaging section 23 into, for example, an 8-bit digital image signal (hereinafter referred to as image data).

The image processing unit 25 performs predetermined image processing such as white balance, γ correction, shading correction and the like on the image data input from the A / D converter 24 in the normal mode, and will be described later in the document mode. After performing predetermined image processing such as background removal, color correction, black area correction, and color data replacement, and performing predetermined compression processing, H
The data is output to the D card 10.

The HD card 10 is a recording medium for recording image data constituting a captured image. The card drive control unit 26 controls the drive of the HD card.

A RAM (Random Access Memory) 27
This is a memory for the control unit 34 to perform processing relating to shooting, color registration, and other functions. ROM (Read Only Memory)
Reference numeral 28 stores data and processing programs necessary for controlling the imaging drive, and data and processing programs necessary for performing various processes such as background removal, color discrimination, black area correction, and false color erasure in a document mode described later. Memory.

The distance measuring section 29 detects a subject distance provided behind the light projecting window 4 and the light receiving window 5. The photometric unit 30 has a light receiving element such as an SPC provided at a position behind the photometric window 3 and detects a subject brightness.

In the color registration mode, the set color normalization calculating section 31 uses the captured color sample image data to
This is for calculating normalized data described later. In the normalized data, the levels of the image data of the R, G, and B color components are Xr, Xg, Xb, and the normalized data is xr, xg, xb.
^{When, xr = Xr 2 / (Xr} 2 + Xg 2 + Xb 2), xg = X
^{g 2 / (Xr 2 + Xg} 2 + Xb 2), xb = Xb 2 / (Xr 2 + Xg 2
+ Xb ² ), and is used for color region division when performing color discrimination based on the coordinate position of the normalized data on the xb-xg plane. The details of the color determination method based on the coordinate position of the normalized data will be described later.

The set color normalization calculation unit 31 extracts image data included in the frame 21 in the photographing screen from the image data of the color sample taken in the color registration mode, and further sets a preset data from the extracted data. Image data of a predetermined level or less. This extraction process is shown in FIG.
1 extracts image data corresponding to a red color sample. Then, the set color normalization calculation unit 31 calculates secondary normalized data using the extracted color sample image data. This calculation result is stored in the RAM 2 via the control unit 34.
7 is stored.

The color area setting calculation section 32 calculates the above-mentioned xb-x
This is to divide a color area when performing color discrimination based on the coordinate position of the normalized data on the g plane. That is, for example, when a chromatic color is distinguished into four colors of blue, green, orange, and red, x
In the color discrimination based on the coordinate position of the normalized data on the b-xg plane, the xb-xg plane is divided into four color regions of blue, green, orange, and red, and the coordinate position of the normalized data is in any color region. The color region setting calculation unit 32
Calculates a boundary line for dividing the xb-xg plane into four color regions of blue, green, orange, and red. The calculation result of this boundary line is also stored in the RAM 27. The setting of the color area will also be described later.

The control section 34 centrally controls the photographing operation of the digital camera 1 and is constituted by a microcomputer. The control unit 34 controls the driving of the imaging unit 23 to capture an optical image of the subject, controls the driving of the image processing unit 25 to perform predetermined image processing on the captured image, and controls the driving of the card driving control unit 26. Under control, the captured image after the image processing is recorded on the HD card.

FIG. 6 is a block diagram showing the image processing (color correction processing) of the image processing section 25 in the document mode.

The image processing unit 25 includes circuit blocks for performing color correction processing in the document mode.
Image memory 251, first noise elimination unit 252, background removal calculation unit 253, γ characteristic setting unit 254, color determination unit 25
5, a black area correction unit 256, a second noise elimination unit 257, and a color replacement unit 258.

In the document mode, as shown in FIG. 7, the image data input to the image processing section 25 is first subjected to a pre-processing for whitening the background (processing A), and then to the pixel data of each pixel position. Normalized data is created for the second power of the image data (R, G, and B image data) (Process B), and color discrimination is performed using the normalized data (Process C). Then, among the image data constituting the characters and the like, the image data whose color has been erroneously determined is corrected, and the image data colored in a chromatic color is corrected among the image data of the black region (process D). ), The image data at each pixel position is replaced with predetermined color data based on the color determination result and output to the HD card 10 (Process E).

The image memory 251 temporarily stores the image data of each of the R, G, and B color components input from the A / D converter 24. The first noise elimination unit 252 includes:
For example, the image data at each pixel position is changed to 8 at the adjacent pixel position.
The noise of the image data input to the image memory 251 is reduced by substituting the average value with the image data.

The background removal operation unit 253 separates the background portion and the character portion of the image data, and converts an area corresponding to the background portion into predetermined white level data. That is, for example, in the case of an image of a subject in which characters or figures are drawn on a whiteboard (hereinafter, referred to as a whiteboard photographed image), as shown in FIG.
5 makes the image of the board portion 351 white uniformly and makes the image of the character portion 352 easy to see. The background removal calculation unit 253 performs the background removal processing by performing γ correction of the image data at each pixel position according to the γ characteristic input from the γ characteristic setting unit 254.

The γ characteristic setting section 254 sets the γ characteristic for the background removal processing. γ characteristic setting unit 25
4 creates a histogram of the input image data,
The γ characteristic is determined based on this histogram. FIG.
Is a typical histogram of a whiteboard photographed image. In such a character image, two large and small peaks U1 and U2 generally occur. The large mountain U1 that occurs in the light area
A small mountain U corresponding to the board portion 351 and generated in the dark area
2 corresponds to a clear character or figure in the character portion 352. The valley portion R between the large mountain U1 and the small mountain U2 is a portion of a dirt on the board portion 351 or a portion of a character or figure having a thin or thin line of the character portion 352.

The γ characteristic setting section 254 is, for example, a large mountain U
Calculate the level w of the image data corresponding to one peak,
As shown in FIG. 10, a γ characteristic in which the level w is set to the white saturation level is determined. FIG. 10 shows the γ characteristic when the image data is 8-bit data, and the level is “255”.
Is a white level, and level “0” is a black level. In the γ characteristic shown in FIG. 3, all image data at the level w or higher are uniformly converted to a white level, and image data lower than the level w are linearly converted at a predetermined conversion ratio (γ value) according to the level. Is converted. In order to emphasize black in the character portion 352, a γ characteristic in which a predetermined low level b is set to a black saturation level may be used. γ
In the characteristics, all the image data at the level b and below are uniformly converted to the black level.

Therefore, when the gamma correction of the image data is performed by the background removal calculating unit 253 using the gamma characteristic (or) set by the gamma characteristic setting unit 254, as shown in FIG.
The image data of the board portion 351 is almost uniformly converted into white data, and the background of the captured image is skipped white.

The color discriminating section 255 discriminates the color of each pixel position. The color determination unit 255 includes a normalization calculation circuit 255a, a luminance calculation circuit 255b, a shape recognition circuit 255c, and a color determination circuit 25 as circuits for performing a color determination process.
5d.

As a color discrimination method using image data of R, G, and B color components, as shown in Japanese Patent Application Laid-Open No. Hei 6-105091, two color components out of three color components are used. There is known a method of selecting and performing color discrimination based on the positions of the normalized data of two color components on the coordinate plane of the normalized data of the selected color component. In this embodiment as well, the color discrimination is performed based on the positions of the normalized data of the two color components. However, the above publication discloses the primary normalized data, that is, the mixing of the image data of the R, G, and B color components. Normalized data x representing the ratio
r = Xr / (Xr + Xg + Xb), xg = Xg / (Xr + Xg +
Xb), whereas a color is classified by using xb = Xb / (Xr + Xg + Xb), the present embodiment, the secondary powers normalized for value data ^{xi = Xi 2 / (Xr 2} + Xg 2 + Xb ² )
(I = r, g, b).

In the present embodiment, the use of the secondary normalized data emphasizes the color classification from the viewpoint of making the characters and the like easy to see in the photographing in the document mode. This is because the use of the normalized data is advantageous for the color discrimination processing.

FIG. 11 shows blue, green, and blue colors in the xb-xg plane when the color is determined using the primary normalized data xb and xg.
FIG. 12 is a diagram showing the color distribution of each system of orange and red, and FIG.
X for color discrimination using the next normalized data xb, xg
It is a figure which shows the color distribution of each system of blue, green, orange, and red in a b-xg plane.

In both figures, (xb, xg) = (1/3,
The 1/3) Q point is a position where it is completely determined to be achromatic when photographed by a camera having an ideal white balance. Also, (xb, xg) = (1, 0), (0,
The points 1) and (0, 0) are positions that are determined to be complete blue, green, and red, respectively. Since an actual camera does not have an ideal white balance, even if a pure blue color is photographed, for example, normalized data (x
(b, xg) does not satisfy (xb, xg) = (1, 0), and is generally located in a certain area biased toward the Q point.

For example, when a color sample created on a whiteboard is photographed with four standard color pens of blue, green, orange and red, normalized data (xb, xg) of each color is as follows:
Each is distributed in the areas indicated by the ellipses Sb, Sg, So, and Sr. Further, a white or black achromatic color is also distributed in a certain region Sw near the point Q.

As is clear from comparison of the ellipses Sb, Sg, So, and Sr in both figures, the use of the second-order normalized data (xb, xg) makes the ellipses Sb, Sg, So, and Sr more flat. Become
It can be seen that the position of the normalized data (xb, xg) generated from the same chromatic image data is far from the ideal achromatic point Q (plotted in the blue region in FIGS. 11 and 12). Compare points P and P '). Therefore, color discrimination can be performed more easily and accurately using secondary normalized data than using primary normalized data.

[0062] normalization arithmetic circuit 255a is, R of each pixel position, G, the image data secondary powers values normalized data xr = Xr ² / by the respective color components ^{B (Xr 2 + Xg 2 +} Xb 2), xg
^{= Xg 2 / (Xr 2 +} Xg 2 + Xb 2), xb = Xb 2 / (Xr 2 +
Xg ² + Xb ² ).

The normalization arithmetic circuit 255a calculates the image data Xr, Xg,
Xb is converted into secondary normalized data xr, xg, xb.

That is, the image data Xr, X
g and Xb are each squared, and the operation results are stored in registers Rr1, Rg1, and Rb1, respectively (# 2). Subsequently, the square values Xr ² , Xg ² , and Xb ² of the image data Xr, Xg, and Xb stored in the registers Rr1, Rg1, and Rb1 are added, and the result is stored in the register SUM (# 4). Subsequently, the square values Xr ² , Xg ² , and Xb ² of the image data Xr, Xg, and Xb stored in the registers Rr1, Rg1, and Rb1 are respectively added to the added value (Xr ² + Xg ² +) stored in the register SUM.
Xb ² ) and divide the result into registers Rr2 and Rg, respectively.
2, stored in Rb2 and the process is terminated (# 6).

When the normalized data (xb, xg) of an order larger than the second order is used, the tendency of the position of the normalized data (xb, xg) on the xb-xg plane to be far from the achromatic point Q is stronger. And the color characteristics of the image data (ie blue, green, orange, or red)
Since but a clearer, as the normalized ^{data, xr = Xr k / (Xr} k + Xg k + Xb k) xg = Xg k / (Xr k + Xg k + Xb k) xb = Xb k / (Xr k + Xg k + Xb k ), K-th (k> 2) normalized data may be used.

In this embodiment, the image data of the R, G, and B color components has been described. However, normalized data X (1), X (2),. Even when (n) is used, normalized data x for a power value of second or higher order
(1) ^k = X (1) ^k / ΣX (i) ^k , x (2) ^k = X (2) ^k / ΣX (i) ^k ,
.. X (n) ^k = X (n) ^k / ΣX (i) ^k (i = 1, 2,... N, k ≧
2) can be used.

In the color discrimination, it is necessary to discriminate between chromatic colors and achromatic colors, such as white and black, as well as chromatic colors. The distinction between chromatic and achromatic white and black can also be made by examining the position of the normalized data in the xg-xb plane as described above.

However, the normalized data generated from the image data of the R, G, and B color components abstracts only the deviation of the R, G, and B color components in the image, and the information of light and dark is discarded. Therefore, if color discrimination is performed using only normalized data, low-luminance black or high-luminance white may be erroneously determined as a chromatic color such as red or blue. That is, for example, if the image data (Xr, Xg, Xb) of the R, G, B color components is (20, 5,
5) Secondary normalized data (x
b, xg) is (0.06, 0.06), and the image data (X
Since it is the same as the second-order normalized data (xb, xg) for the red pixel position of (r, Xg, Xb) = (200, 50, 50), black is erroneously determined to be red.

Therefore, in order to prevent such erroneous determination, in this embodiment, color determination (including achromatic color determination) is performed using the luminance data at each pixel position and the above-described normalized data. Like that. Further, in order to reduce the influence of the line width on the color discrimination, as described later, the color discrimination is performed in consideration of a geometric shape such as a character or a figure.

The brightness calculation circuit 255b calculates brightness data at each pixel position used in the above-described color discrimination processing. The luminance calculation circuit 255b calculates the average value Y (= (Xr +) of the image data Xr, Xg, Xb of the R, G, B color components, for example.
Xg + Xb) / 3) is calculated as luminance data for color determination at each pixel position.

The shape recognizing circuit 255c recognizes whether an area constituting a character, a figure or the like is linear or planar. The shape recognition circuit 255c corrects the level of the image data in each block of 5 × 5 pixels or 7 × 7 pixels so that the maximum value becomes “255”, and then “0” and “255”. As shown in FIG. 14, a pattern α of a thin line portion and a pattern α of a thick line portion are obtained by performing a pattern matching by a square error calculation or the like with a template having image data of several levels or image data of several levels of “0” to “255”. Is detected. This detection result is input to the color determination circuit 255d. Note that the shape determination may be performed by high-frequency component detection processing instead of pattern matching.

The color discrimination circuit 255 d is
1 based on the normalized data (xb, xg) calculated in 55a, the brightness data Y calculated in the brightness calculation circuit 255b, and the pattern result recognized in the shape recognition circuit 255c.
The determination of the color of each pixel position (for example, the color determination of white, black, blue, green, orange, and red) is performed using a predetermined threshold profile shown in FIGS.

FIG. 15 is a diagram showing a first embodiment of a threshold profile for performing color discrimination between a chromatic color and an achromatic color based on luminance data for color discrimination. 16 to FIG.
8 is a modified example of a threshold profile for performing color discrimination between a chromatic color and an achromatic color based on luminance data for color discrimination, and is a diagram illustrating second to fourth embodiments, respectively.

15 to 18, the horizontal axis represents xb-x
The distance information from the achromatic color point Q on the g plane is shown, and the vertical axis shows the luminance level for color discrimination.

It should be noted that the distance information corresponds to the point Q on the xb-xg plane.
Does not indicate the actual distance from the position of the normalized data (xb, xg) to the achromatic point (1/3, 1/3, 1/3) in the xr-xg-xb space. Xb-xg for closed surface
Equidistant line (achromatic point Q (1/3, 1
/ 3) is the distance information for specifying the closed curve including).

Here, the distance information will be briefly described. If arbitrary normalized data (xb, xg) on an equidistant line on the xb-xg plane is treated as a group of data at the same distance, the equidistant line is If it is a circle centered on the achromatic point Q (1/3, 1/3), the actual distance from the point Q on the xb-xg plane to the position of the normalized data (xb, xg) as distance information Can be adopted.

However, the equidistant line is xr-x as described above.
A closed surface equidistant from the achromatic point (1/3, 1/3, 1/3) in the g-xb space is projected onto the xb-xg plane, and the achromatic point Q (1/3, 1 / Since the curve is a concentric closed curve that is not a circle centered on 3), the actual distance from the point Q to the position of the normalized data (xb, xg) cannot be used as distance information.

Therefore, in the present embodiment, since the same distance information is given to the normalized data (xb, xg) on the equidistant line, the normalized data (xb, xg) and the equidistant line to which it belongs belong to a pair. Therefore, the distance information defining the position of the equidistant line on the xb-xg plane is set as the distance information of the normalized data (xb, xg) on the equidistant line.

Given the normalized data (xb, xg), to determine the distance information for the normalized data (xb, xg), the equidistant line to which the normalized data (xb, xg) belongs is determined by You need to decide. However, xr-
A plurality of equidistant lines formed by projecting a plurality of equidistant surfaces formed concentrically around an achromatic point (1/3, 1/3, 1/3) in the xg-xb space onto the xb-xg plane In the normalized data (x
It is not necessary to determine the equidistant line to which (b, xg) belongs. Because the distance information of the normalized data (xb, xg) is used for color discrimination by relative comparison with a predetermined threshold value, the achromatic point Q (1/3, 1 / This is because the position of the normalized data (xb, xg) from 3) may be specified.

If the relative distance information is determined from the equidistant line for the given normalized data (xb, xg), the normalized data (xb, xg)
And a plurality of normalized data (x
b, xg) can be used.

On the other hand, in the xr-xb-xg space, six normalized data (xr, xb, x
g), (xb, xg, xr), ... are achromatic points (1/3, 1/3,
1/3) equidistant from the
Since the points projected on the xg plane form equidistant lines on the xb-xg plane, the shape that can be adopted as the equidistant lines is at least the given normalized data (xb, xg) and the components of the normalized data. Other 5 generated by exchanging
Pieces of normalized data (xr, xb), (xr, xg), (x
b, xr), (xg, xr), (xg, xb).

Accordingly, on the xb-xg plane, six components on the xb-xg plane generated by exchanging the three components with each other are shown.
Pieces of normalized data (xb, xg), (xr, xb), (x
r, xg), (xb, xr), (xg, xr), (xg, xb)
The arbitrary shape passing through can be a specific shape of the equidistant line.

FIG. 19 is a diagram showing an example of the positional relationship on the xb-xg plane of six normalized data obtained by exchanging components.

P1 (1/2, 1/3) and P2 (1/1) shown in FIG.
2, 1/6), P3 (1/3, 1/6), P1 '(1/3, 1/2), P
Each point of 2 '(1/6, 1/2) and P3' (1/6, 1/3) is the point P
1, normalized data xr (= 1/6), xg (= 1/2),
The five normalized data generated by exchanging the components of xb (= 1/3) are plotted on the xb-xg plane. As is clear from the figure, the points P1 ', P2', P
3 'is a line symmetrical point with respect to a straight line L (a straight line connecting the origin and the point Q) of the points P1, P2 and P3. The points P3 and P3 'are symmetrical points with respect to the straight line L' of the points P1 and P1 '(a straight line orthogonal to the straight line L passing through the point Q). Such points P1 to P3, P1 '
The symmetry of P3 'is based on the fact that the components having the relationship of xr + xb + xg = 1 are exchanged with each other to determine the normalized data of each point.

Therefore, the equidistant line to which point P1 belongs is the point P
1 to P3, P1 'to P3', and straight lines L, L '
Must be symmetrical with respect to the shape, such as an ellipse C1 shown in FIG.
An isosceles triangle C2 shown in FIG. 21 or a hexagon C3 shown in FIG. 22 can be considered.

When the shape of the equidistant line is specified, if the method of calculating the distance from the achromatic point Q (1/3, 1/3) is determined for the equidistant line, the normalized data (Xb, x
calculating the equidistant line of the specific shape to which the normalized data (xb, xg) belongs from g), and further calculating the distance from the achromatic point Q (1/3, 1/3) for the equidistant line By
Distance information for the given normalized data (xb, xg) can be determined.

In the examples shown in FIGS. 20 to 22, the distance from the achromatic point Q (1/3, 1/3) of each equidistant line, for example, a straight line L
The distance d between the point P4 and the point Q at the point of intersection of the equidistant lines C1, C2 and C3 can be taken. This distance d is the distance from the center Q of the ellipse C1 to the closest point (１ ／ of the short axis length) when the ellipse C1 is adopted as the closed curve, and when the isosceles triangle C2 is adopted as the closed curve, This is the distance from the point Q to the midpoint of the hypotenuse.

Although the equidistant lines C2 and C3 have different shapes, the oblique sides passing through the points P3 and P3 'are the same, so that the distance d from the point Q is the same. Therefore, when considering the distance information, the equidistant lines C2 and C3 are practically the same.

In this embodiment, the normalized data (xb,
An operation table for directly calculating the distance information d from xg) is provided. When the normalized data (xb, xg) is calculated, the distance information d is directly calculated using this operation table.

Returning to FIG. 15, the threshold profile shown in FIG. 15 is such that an area in the distance information d1 from the achromatic position is defined as an achromatic area Sw, and in this area Sw, the threshold value for determining white and black is a fixed value Y2. Outside the achromatic region, the threshold value for determining chromatic color and black is fixed at Y1, and the threshold value for determining chromatic color and white is fixed at Y3 (Y1 <Y2 <Y3).

In this threshold profile, when the position of the normalized data (xb, xg) is located inside the equidistant line of the distance information d1 (d ≦ d1), the luminance level Y is compared with the threshold Y2. If Y ≧ Y2, it is determined that the color is white,
If Y <Y2, it is determined to be black. When the position of the normalized data (xb, xg) is located outside the equidistant line of the distance information d1 (d> d1), the luminance level Y
Are compared with the threshold values Y1 and Y3. If Y> Y3, the color is determined to be white. If Y <Y1, the color is determined to be black.
If ≧ Y ≧ Y1, it is determined to be chromatic. In the case of a chromatic color, the above-described normalized data (xb, xg)
, Four colors, for example, blue, green, orange, and red are determined.

The threshold profile shown in FIG.
, The threshold value for determining the chromatic color and white or black when the position of the normalized data (xb, xg) is outside the equidistant line of the distance information d1 is Y1, Y3 to Y1, Y3, Y4.
Y5 (Y3>Y4>Y1>Y5> Y2) is increased in two steps. In this threshold profile,
When the position of the normalized data (xb, xg) is between the equidistant line of the distance information d1 and the equidistant line of the distance information d2 (d2
>D> d1) and if Y> Y4, it is determined to be white and Y
If <Y5, it is determined to be black, and if Y4 ≧ Y ≧ Y5, it is determined to be chromatic. Also, normalized data (x
b, xg) is the equidistant line of the distance information d2 and the distance information d
3 (d3>d> d2), Y
> Y3, it is determined to be white, and if Y <Y5,
The color is determined to be black, and if Y3 ≧ Y ≧ Y5, the color is determined to be chromatic. Further, when the position of the normalized data (xb, xg) is between the equidistant line of the distance information d2 and the equidistant line of the distance information d3 (d> d3), if Y> Y3, it is determined to be white. If Y <Y2, black is determined and Y3 ≧ Y
If ≧ Y2, it is determined to be a chromatic color.

The threshold profile shown in FIG.
In, when the position of the normalized data (xb, xg) is between the equidistant line of the distance information d1 and the equidistant line of the distance information d2, the threshold for determining the chromatic color and white linearly changes from Y4 to Y3. When the position of the normalized data (xb, xg) is between the equidistant line of the distance information d1 and the equidistant line of the distance information d3, the chromatic / black discrimination threshold is linearly changed from Y5 to Y2. It was made.

The threshold profile shown in FIG.
In FIG. 16, when the position of the normalized data (xb, xg) is between the equidistant line of the distance information d1 and the equidistant line of the distance information d2, the chromatic and white discrimination thresholds are non-linear from Y4 to Y3. (Monotone increase), and the chromatic / black discrimination threshold when the position of the normalized data (xb, xg) is between the equidistant line of the distance information d1 and the equidistant line of the distance information d3 is Y5. To Y2 in a non-linear (monotonically decreasing) manner.

FIGS. 15 to 18 show typical examples of threshold profiles, and other threshold profiles can be arbitrarily adopted.

In the threshold profiles shown in FIGS. 15 to 18, normalized data (xb, xb) on the xb-xg plane
When the four colors of blue, green, orange, and red are determined from the position g), ellipses Sb, Sg, So, and Sr are used as shown in FIG.
And dividing the triangle xg-xb plane into four regions Ab, Ag, Ao, and Ar by providing dividing lines m1, m2, m3, and m4, and setting the position of the normalized data (xb, xg) to four Area A
Color discrimination is performed depending on which of b, Ag, Ao, and Ar (excluding the area Sw). For example, if the coordinates (xb, xg) of the normalized data are within the region Ab, the pixel position is determined to be a blue-based color, and if the coordinate is within the region Ag, the pixel position is determined to be a green-based color. Is determined.

The division lines m1, m2, m3, and m4 are set by the color area setting calculation unit 32, and the points Pb and Pb in FIG.
g, Po, and Pr are color-registered or preset blue, green, orange, and red image data (default values)
For example, assuming the coordinate position of the regular data of ∠PbQPg,
The bisector of ∠PgQPo, ∠PoQPr, ∠PrQPb or △ P
It is set as a middle line connecting the point Q of bQPg, △ PgQPo, △ PoQPr, △ PrQPb and the midpoint of the opposite side. As described above, in the digital camera 1 according to the present embodiment,
Since the color to be discriminated can be arbitrarily set in the range of four colors of blue, green, orange and red, for example, as shown in FIG. "OF
If "F" is set, division lines m1, m2, and m3 are set for the three colors of blue, green, and red set to "ON", and color discrimination is performed for the three colors. in this case,
The dividing line m3 is set as a bisector of ∠PgQPr or a middle line connecting the point Q of △ PgQPr and the midpoint of the opposite side.

The color discriminating circuit 225d includes the shape recognizing circuit 25.
If the pattern result input from the shape recognition circuit 255c changes, the color determination is performed using a preset threshold profile in a fixed manner. As explained, the threshold profile is changed as a whole according to the change of the pattern result.
Alternatively, the color of the corresponding region is determined using the threshold profile corrected so as to be partially moved up and down. This is because the threshold profile is entirely or partially raised or lowered according to the width of the line constituting a character or the like, so that an image having a preferable line width can be obtained regardless of a change in the line width. It is.

FIG. 24 is a diagram showing an example of a change in density level based on a difference in line width.

This figure shows the density level of the image data on the straight line M of the character portion 36 of the numeral "1" surrounded by the square of the whiteboard photographed image. Character part 36
The graph drawn on top of indicates the concentration level.
The portions A and C of the straight line M are portions corresponding to the square line width (thin line portion), and the portion B is the portion corresponding to the line width of the numeral “1” (thick line portion). A and C are thin line portions,
Since B is a thick line portion, the density levels of A and C are higher than the density level of B. That is, the thin line portion is whiter than the thick line portion.

FIG. 25 shows black and white discrimination of image data on the straight line M in FIG. 24 using a threshold profile.

For example, assuming that a threshold profile having the form shown in FIG. 18 is used, FIG. 26A shows a case where black and white discrimination is performed using the threshold profile having a high threshold level shown in FIG. 26, and FIG. In FIG. 26, black-and-white discrimination is performed by using a threshold profile having a low threshold level in FIG. FIG. 3C shows a threshold profile according to the line width.
Are used for black and white determination.

When a threshold profile having a high threshold level is used, the thick line portion B has a suitable line width, but the thin line portions A and C are higher than the threshold level Y1, so that they are erroneously determined to be white and have a rectangular figure. Will fly away. On the other hand, when a threshold profile having a low threshold level is used, the thin line portions A and C have an appropriate line width, but the thick line portion B
Means that the image data of the outline portion also becomes lower than the threshold level Y1 ', and the line width becomes unnecessarily thick.
As described above, when the line width is large or small, a fixed threshold profile is often inappropriate for one line width but inappropriate for the other line width. Irrespective of this, it is difficult to perform appropriate black-and-white discrimination.

Therefore, in the present embodiment, as shown in FIG. 10C, the line width is determined, and the threshold profile is switched according to the result of the determination so that a suitable line width can be obtained. I have. Therefore, the color determination circuit 225d
When the determination result input from the shape recognition circuit 255c is the pattern α, black-and-white determination is performed using the threshold profile, and when the determination result is the pattern β, the threshold profile is switched to the threshold profile to perform black-white determination. .

As described above, the correction of the threshold profile in the black-and-white determination may be performed by switching a plurality of preset threshold profiles in accordance with the determination result of the line width. The profile may be entirely or partially shifted up and down by a predetermined level according to the line width determination result.

Returning to FIG. 6, the black area correction unit 256 corrects a black area containing a chromatic color to a single black color not containing a chromatic color by erroneously determining the color.

When performing color discrimination using the threshold profiles shown in FIGS. 15 to 18, if the distance information d1, which is the threshold for discriminating between achromatic colors and chromatic colors, is set high, chromatic colors with relatively dark colors (ie, A chromatic color with a low luminance is likely to be erroneously determined to be black, and a chromatic color with a relatively pale color (ie, a chromatic color having a high luminance) is likely to be erroneously determined to be white. Therefore, in order to prevent such erroneous determination, it is desirable to set the distance information d1 lower.

However, when a threshold profile in which the distance information d1 is set to a low value is used, black is erroneously determined to be chromatic, and as shown in FIG. May be an area. The reason that the black area tends to become a mottled area is that the level of the black image data is low, so that the normalized data xb and xg are xb / xg =
1 is rare, and the position of the normalized data (xb, xg) is likely to be outside the achromatic region Sw.
When the average of the image data is black in a certain large area, there is a pixel position in which black is erroneously determined as red, green, orange, or the like, for example, near a pixel position in which black is erroneously determined to be blue. It is based on the high possibility.

When a relatively large black area constituting a character, a figure, or the like becomes a mottled area, the image becomes extremely hard to see. The black area correction unit 256 converts the mottled area into a black area,
It reduces the image quality defect caused by the spot pattern, and includes a same-color pixel number counting circuit 256a, a dispersion degree calculation circuit 256b, and a black discrimination circuit 256c.

The same color pixel count circuit 256a is a circuit for counting the number of pixels of the same color in a block. When the block size is 5 × 5 pixels and the color distribution at each pixel position is as shown in FIG. 28, for example, the same-color pixel number counting circuit 256a performs block processing for each color of red, blue, orange, green, and black. Is counted. Red and blue,
The number of pixels of each color of orange, green, and black is represented by Nr, Nb, No, Ng, Nb.
Assuming that k and Nt are the sum of these, in the example of FIG. 9A, Nr = 17, Nb = 2, No = 3, Ng = 1, and Nbk =
2, Nt = 25 is counted, and in the example of FIG.
Nr = 4, Nb = 4, No = 5, Ng = 6, Nbk = 4, Nt
= 23 is counted. In the example of (b), two white pixels are included in the block, but these pixels are ignored in the black area correction processing.

The dispersion degree calculation circuit 256b includes red, blue, orange,
The number of pixels Nr, Nb, No, Ng, Nbk of each color of green and black and the total Nt thereof are used to calculate the degree of dispersion p of the color of each block. The degree of dispersion p is p = (number of pixels of colors other than white) / Σ (number of pixels of each color) ² = Nt / (Nr ² + Nb ² +
No ² + Ng ² + Nbk ² ). The degree of dispersion p is the reciprocal of the variance operation formula. FIG.
In the example of (a), p = 23 / (4 ² +4 ² +5 ² +6 ² +4
² ) = 0.111 is calculated, and in the example of FIG.
= 25 / (17 ² + 2 ² + 3 ² + 1 ² + 2 ²⁾ = 0.081
Is calculated.

The black discrimination circuit 256 c is
It is determined whether or not the pixel at the center position of each block is black using the degree of dispersion p calculated in 56b.
That is, the black determination circuit 256c compares the preset threshold value K1 with the degree of dispersion p and determines that the color is black if p> K1, and determines that the color is chromatic if p ≦ K1. For example, if K1 = 0.1, the example of FIG. 28A is determined to be black, and the example of FIG. 28B is determined to be chromatic. In the example of FIG. 28B, since the number of pixels of each color is overwhelmingly red, the pixel at the center can be determined to be “red”.

FIG. 29 is a flowchart showing the procedure of the correction process of the black area correction unit 256.

The color discrimination result output from the color discrimination unit 255 is stored in a memory (not shown) in the black area correction unit 256.
The correction processing of the black area is performed according to the flowchart of FIG. That is, first, the color determination data for one block (5 × 5 pixels) at the upper left corner is read from the memory (#
10), the number of pixels of each color other than white included in the block and the total number thereof are counted (# 12). Assuming that the discrimination colors are red, blue, orange, green, and black, the number of pixels Nr, N
b, No, Ng, Nbk and their total number Nt are counted.

Subsequently, the degree of dispersion p (= Nt / (Nr ² + Nb ²⁾
+ No ² + Ng ² + Nbk ² )) is calculated (# 14), and the variance p is compared with a predetermined threshold K1 to determine whether or not the pixel at the center of the block is black (# 16). .
If p> K1 (YES in # 16), it is stored that the color determination data at the pixel position at the center of the block (the pixel position in the third row and third column of the block) is changed to black data (# 18). ). On the other hand, if p ≦ K1 (YE in # 16)
S), the data change instruction is not issued.

Subsequently, it is determined whether or not the above processing for the entire color determination data has been completed (# 20). If the processing has not been completed (NO in # 20), the read position of the block is set to one pixel. Shifting in the horizontal direction (# 22), the process returns to step # 10, and the above-described black determination processing is performed (# 10).
~ # 20). Hereinafter, while the readout position of the block is shifted by one pixel in the raster direction, black determination processing is performed on the color determination data at each pixel position (loop of # 10 to # 22), and black determination processing is performed on the entire color determination data. Is completed (YES in # 20), the color determination data at the pixel position instructed to change the color determination data to black data is changed to black data (# 24), and the process ends.

In the present embodiment, the size of one block in the black area correction processing is set to 5 × 5 pixels. However, the size of one block is not limited to this, and processing accuracy and processing speed are reduced. Any appropriate size can be set in consideration of this. Further, although the degree of dispersion p is defined as a measure of the degree of dispersion of colors, other arithmetic expressions may be defined as long as they can represent color dispersion. For example, standard deviation σ =
1 / √ (p) may be used.

The colors of blue, green, orange, red, and black are represented by, for example, (0, 1, 0, 0), (0, 0, 1, 0),
(0,0,0,1), (1,0,0,0), (1,1,
When a vector is expressed as in (1, 1), the reciprocal of the vector length of the average of the vectors in the block may indicate the degree of dispersion. Note that the vector expression may be independent of each color or may be one in which information such as "red and orange are similar colors" is added. Further, the present invention is not limited to four dimensions, and any dimension can be considered.

As a simple method, the number of colors present at eight pixel positions adjacent to the center position of the block and the degree of the spot may be determined based on the number of different colors.

The second noise elimination section 257 corrects a chromatic color or a white color (misidentified color) included in the black area where the color is determined to black. That is, the black area correction unit 2
Even if the process of correcting the mottled pattern area into a black area by the step 56, the chromatic color may still remain in an island shape in the black area, the second noise canceling unit 257 converts this false color to black. And erase it.

FIG. 30 is a diagram showing an example of a state in which chromatic colors remain in an island shape in a black area such as a character.

FIG. 27 shows a case where chromatic pixels 37 and 38 remain in the black region forming the numeral “2”. Is corrected to black, and the false color is eliminated. Note that black correction is not performed on the pixel 38 located at the outline position of the black area forming the number “2”. When a color other than black is mixed in an island shape in the black area, the color is conspicuous and the character is hard to see, but when a color other than black is mixed at the border, the adverse effect of the false color is Since the number is small, the correction process is performed only inside the black region in consideration of the processing speed delay due to the correction process for the black region. It should be noted that black correction may also be performed on a false color at the outline position of the black area.

FIG. 31 is a flow chart showing the processing procedure of the false color erasure of the second noise erasing section 257.

In the flowchart shown in FIG. 32, as shown in FIG. 32, the colors of the center position G _A and eight adjacent pixel positions are checked, and if the same color exceeds a predetermined number K2, the center pixel G _A is determined. The color of the position G _A is changed to a color exceeding the predetermined number K2. Figure 32 is a predetermined number K2 obtained by six, the seven pixels excluding the center and the lower right corner is in the black, show that the color of the pixel G _A center position is changed to black ing.

The color determination data output from the black area correction section 256 is stored in a memory (not shown) in the second noise elimination section 257, and the black area correction processing is performed according to the flowchart of FIG. That is, first, the color discrimination data of one block (3 × 3 pixels) for the upper left is read from the memory (# 30), and the number of pixels of each color included in the block is counted (# 32). .

Subsequently, the maximum number of pixels Nmax is compared with a predetermined threshold value K2, and it is determined whether or not Nmax> K2 (# 34). If Nmax> K2 (YE in # 34)
S), the pixel position at the center of the block (2 rows 2
The change of the color determination data of the pixel position in the column) to the data of the color of the number of pixels Nmax is stored (# 36). on the other hand,
If Nmax ≦ K2 (YES in # 34), the above-described data change instruction is not performed.

Subsequently, it is determined whether or not the false color correction processing for the entire color determination data has been completed (# 38). If the processing has not been completed (NO in # 38), the reading position of the block is set to one pixel. The image is shifted in the horizontal direction by an amount (# 40), and the process returns to step # 30, where the above-described false color correction processing is performed (# 30 to # 38). Hereinafter, a false color correction process is performed on the color determination data at each pixel position while shifting the block reading position by one pixel in the raster direction (# 30 to # 4).
When the false color correction processing for the entire color determination data is completed (YES in # 38), the color determination data at the pixel position instructed to change the color determination data to the data of the number of pixels Nmax is obtained. The data is changed to the designated color data (# 42), and the process ends.

Returning to FIG. 6, the color replacing section 258 converts the image data at each pixel position into a color determining section 255 and a black area correcting section 25.
6 and corresponding color image data (image data of R, G, B color components) set in advance based on the color data set by the second noise canceling unit 257 or set by color registration. Replace it. Then, the color replacement unit 258
The image data replaced by is output to the HD card and recorded (see FIG. 5).

Next, the photographing operation of the digital camera 1 will be described with reference to the flowcharts of FIGS.

FIG. 33 is a flowchart showing a photographing operation procedure in the document mode.

When the main switch 11 is turned on (# 5)
0, YES), the digital camera 1 is activated, and is in a photographable state. First, it is determined whether or not the mode change switch 16 is set to the shooting mode (# 52). If the shooting mode is not set, that is, if the color registration mode is set (NO in # 52). The color registration process is performed according to the flowchart shown in FIG. 34 (# 54).
If the shooting mode has been set (YES in # 52), the flow shifts to step # 54, where shooting processing is performed.

When the process proceeds to the color registration process, first, the color designated by the color selection button 18 is set as a color to be registered (# 80). In this color setting process, the color to be registered is determined and set based on the display position of the bar display 171. Subsequently, whether or not the set color has not been registered is determined based on the presence or absence of the mark 172 (# 82).
If not registered (NO in # 82), it is further determined whether or not an instruction to delete registered color data is given by the default button 20 (# 84).

If deletion of the registered color data is instructed (YES in # 84), the secondary normalized data of the registered color stored in the RAM 27 is rewritten to the preset default value of the color. (# 84), Step # 8
Return to 2. On the other hand, if the deletion of the registered color data has not been instructed (NO in # 84), the process proceeds to step # 88.

If it is determined in step # 82 that the set color has not been registered (YES in # 82), it is determined whether or not imaging of a color sample (see FIG. 4) has been instructed by the release operation of the shutter button 9. Then, if the imaging is instructed (# 88: YES), the photometric unit 30 detects the subject luminance (# 90), and the distance measuring unit 29 detects the subject distance (# 92). The focus is adjusted based on the subject distance (# 94). Further, an exposure control value (aperture value and CCD integration time) is set using the detected subject brightness (# 96).

Subsequently, an image of a color sample is taken based on the set exposure control value (# 98). From the image data captured by the imaging unit 23, the image data in the frame 21 at the center of the shooting screen is extracted by the set color normalization calculation unit 31, and the image data of only the color sample portion is extracted from the image data. (# 100). Then, secondary normalized data is calculated for the color sample image data (# 102), and the calculation result is stored in a predetermined storage area of the RAM 27 (# 104).

RAM 27 for secondary normalized data of color sample
When the storage (color registration) is completed, it is subsequently determined whether or not the use of the color is instructed by operating the set button 19 (# 106). If the use of the color is instructed (# 1)
(YES in 06), the color use in the color discrimination of the registered color is set (# 108), and the routine returns. At this time, the color registration confirmation display section 17 displays the color use “on”. If the use of color is not instructed (N in # 106)
O), the use of the color in the color discrimination of the registered color is released (# 1)
10) Return. At this time, the color registration confirmation display section 1
At 7, the display of color use "off" is performed.

On the other hand, if imaging has not been instructed in step # 88 (NO in # 88), steps # 90 to # 1
The step jumps to step # 106 without performing the above-described process of registering the secondary normalized data of the color sample.

As described above, in the color registration mode,
When the default button 20 is operated, the image data (R, R,
The image data of the G and B color components) and the secondary normalized data are deleted, and instead, preset image data and secondary normalized data (default values) are set.
In addition, when the set button 19 is operated, setting or cancellation of color use is performed for the set color. Further, when a color sample is photographed regardless of whether it is unregistered or registered, secondary normalized data of the photographed color sample is calculated, and the calculated value and the captured image data are newly or Registered renewally.

Returning to FIG. 33, if the photographing mode is set in step 52 (YES in # 52), subsequently, it is determined whether or not photographing is instructed by the release operation of the shutter button 9 (#). 56) If imaging is not instructed (NO in # 56), the process returns to step # 52, and if imaging is instructed (YES in # 56), is the color correction correction switch 13 further set to ON? It is determined whether or not the document mode is set (# 58).

If the document mode is not set, that is, if the normal mode is set (#
(NO at 58), a normal photographing process is performed. That is, the subject brightness and the subject distance are detected, and after the focus adjustment and the exposure control value setting are performed based on these detected values, the subject is photographed. The image data captured by the imaging unit 23 is subjected to predetermined image processing (image processing for obtaining high image quality of depiction) by the image processing unit 25, compressed, and recorded on the HD card 10. .

On the other hand, if the document mode is set (YES in # 58), first, the density setting switch 15
The density data (dark / light data) is taken in according to the set position (# 62). Subsequently, the image processing unit 25 is set to image processing in the document mode, and the density data is set in the color determination unit 255 (# 64).

Subsequently, the luminance of the subject is detected by the photometric unit 30 (# 66), the subject distance is detected by the distance measuring unit 29 (# 68), and the focus is adjusted based on the subject distance (# 66). 70). Also, an exposure control value (aperture value and CCD integration time) is set using the detected subject brightness (# 72).

Subsequently, an image of a subject is taken by the image pickup section 23 based on the set exposure control value (# 74). The image data taken in the image pickup operation is transferred to the image processing section 25 in accordance with the flowchart shown in FIG. After the predetermined color correction processing is performed (# 76), it is recorded on the HD card 10 (# 78). And this ends one shot,
The flow returns to step # 52 to perform the next photographing.

When the process proceeds to the color correction processing, first, the γ characteristic for the background removal processing is set by the γ characteristic setting section 254 using the image data stored in the image memory 251 (# 120). Subsequently, after the image data stored in the image memory 251 is subjected to noise reduction processing by the first noise elimination unit 252, the background removal processing unit 253 performs γ correction using the above γ characteristic, thereby performing background removal processing. (Process for uniformly converting a white background portion to white) is performed (# 122).

Subsequently, secondary normalization data for color discrimination processing is calculated by the normalization calculation circuit 255a of the color discrimination unit 255 using the image data stored in the image memory 251 (# 124). Further, the luminance data for color determination is calculated for each pixel position using the image data stored in the image memory 251 with the brightness calculation color 255b of the color determination unit 255 (# 126). Further, the shape recognizing circuit 255c of the color discriminating unit 255 uses the image data stored in the image memory 251 to perform a shape recognizing process (pattern recognizing process by pattern matching) of characters, figures, and the like on a predetermined block basis (# 128). The color discrimination circuit 255d of the color discrimination unit 255 uses the above-described color discrimination method by using a preset threshold profile (see FIGS. 15 to 18) based on secondary normalized data, luminance data, and a shape recognition result. Thus, the color of each pixel position is determined (# 128).

Subsequently, the black area correction unit 256 performs a process of correcting a spot pattern on the color determination data (# 132), and the second noise elimination unit 257 performs an island-like process on the color determination data. After the process of erasing the false color that has occurred is performed (# 257), the color replacement unit 258 converts the image data at each pixel position (image data at the time of shooting) into predetermined color data based on the color determination data. (Image data of R, G, B color components set in advance (default value) or image data of R, G, B color components registered in the registration process) (# 136), thereby The correction processing ends, and the process returns.

In the above embodiment, a digital camera has been described as an example. However, the present invention can be applied to other color image processing devices such as a color discriminating process by a computer.

[0148]

As described above, according to the present invention,
Color data is extracted in units of a predetermined number of pixels from the color determination result of each pixel position determined using the color component image data of the color image, and the color distribution of the extracted area is When the color variance is determined to be large based on the degree of variance, the color data at the pixel position at the center of the extraction area is replaced with black data. The chromatic color generated as noise in the black area due to the erroneous determination can be suitably removed.

Further, according to the present invention, when it is determined that the color variance is large based on the degree of variance, the color data of the entire extraction area is replaced with black data. Processing can be sped up.

[Brief description of the drawings]

FIG. 1 is a perspective view showing an appearance of a digital camera provided with a color discriminating apparatus according to an embodiment of the present invention.

FIG. 2 is a rear view of a digital camera provided with the color determination device according to the present invention.

FIG. 3 is a diagram illustrating an example of a display of a color registration confirmation display unit.

FIG. 4 is a diagram illustrating a state in which a frame indicating a color data capture range is displayed in a finder in a color registration mode.

FIG. 5 is a block diagram illustrating image processing of a captured image in a document mode of the digital camera according to the present invention.

FIG. 6 is a block diagram illustrating image processing performed by an image processing unit in a document mode.

FIG. 7 is a diagram illustrating an outline of an image processing procedure in a document mode.

FIG. 8 is a diagram illustrating the details of a background removal process.

FIG. 9 is a diagram showing a typical example of a histogram of image data constituting a whiteboard photographed image.

FIG. 10 is a diagram illustrating an example of a γ characteristic set for background removal.

FIG. 11 is a diagram showing the color distribution of each of the blue, green, orange, and red systems on the xb-xg plane when color discrimination is performed using primary normalized data xb, xg.

FIG. 12 is a diagram showing the color distribution of each of the blue, green, orange, and red systems on the xb-xg plane when performing color discrimination using the second-order normalized data xb, xg.

FIG. 13 is a flowchart for converting image data Xr, Xg, and Xb into quadratic normalized data xr, xg, and xb.

FIG. 14 is a diagram illustrating a method of detecting a thin line portion and a thick line portion by pattern matching.

FIG. 15 is a diagram showing a first embodiment of a threshold profile for performing color discrimination between a chromatic color and an achromatic color based on luminance data for color discrimination.

FIG. 16 is a diagram illustrating a second embodiment of a threshold profile for performing color discrimination between a chromatic color and an achromatic color based on luminance data for color discrimination.

FIG. 17 is a diagram illustrating a third embodiment of a threshold profile for performing color discrimination between a chromatic color and an achromatic color based on luminance data for color discrimination.

FIG. 18 is a diagram illustrating a fourth embodiment of a threshold profile for performing color discrimination between a chromatic color and an achromatic color based on luminance data for color discrimination.

FIG. 19 is a diagram showing an example of a positional relationship on the xb-xg plane of six normalized data in which components are interchanged.

FIG. 20 is a diagram in which the shape of an equidistant line generated by six normalized data in which components are exchanged is an ellipse.

FIG. 21 is a diagram in which the shape of an equidistant line generated by six normalized data in which components are interchanged is an isosceles triangle.

FIG. 22 is a diagram showing hexagonal shapes of equidistant lines generated by six normalized data in which components are interchanged.

FIG. 23 is a diagram showing a state where an xb-xg plane is divided into blue, green, orange, and red color regions.

FIG. 24 is a diagram showing a relationship between a line width of a character or the like and a luminance level.

25A and 25B show the result of performing black and white discrimination on portions having different line widths such as characters using a threshold profile. FIG. 25A shows a case where a threshold profile with a high threshold level is used, and FIG. (C) shows a determination result when a threshold profile having a low threshold level is used, and (c) shows a determination result when the threshold level is changed according to the line width.

FIG. 26 is a diagram illustrating a correction example of a threshold profile corrected according to a line width.

FIG. 27 is a diagram illustrating an example of a case where a color discrimination result is a mottled pattern.

28A and 28B are diagrams illustrating calculation examples of the degree of dispersion, wherein FIG. 28A is a diagram illustrating the degree of dispersion of a mottled region, and FIG. 28B is a diagram illustrating the degree of dispersion of a red region.

FIG. 29 is a flowchart illustrating a procedure of a correction process of a black area correction unit.

FIG. 30 is a diagram illustrating an example of a state in which chromatic colors remain in an island shape in a black region such as a character.

FIG. 31 is a flowchart illustrating a processing procedure of false color erasure of a second noise erasing unit.

FIG. 32 is a diagram illustrating a method of correcting the color of the center pixel using the color determination results of eight adjacent pixel positions.

FIG. 33 is a flowchart showing a shooting operation procedure in a document mode of the digital camera according to the present invention.

FIG. 34 is a flowchart illustrating a procedure of a color registration process.

FIG. 35 is a flowchart illustrating a processing procedure of a color correction process.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Digital camera 2 Photographing lens 3 Metering window 4 Light-emitting window 5 Light-receiving window 6 Viewfinder objective window 7 Card slot 8 Card eject button 9 Shutter button 10 HD card 11 Main switch 12 Viewfinder eyepiece window 13 Color correction switch 14 Noise correction switch 15 Density setting switch 16 Mode change switch 17 Color registration confirmation display section 18 Color selection button 19 Set button 20 Default button 21 Frame 22 Color sample 23 Imaging section 24 A / D converter 25 Image processing section (color discriminating device) 251 Image memory 252 1st noise elimination section 253 background removal calculation section 254 γ characteristic setting section 255 color discrimination section (color discrimination means) 255a normalization calculation circuit 255b luminance calculation circuit 255c shape recognition circuit 255d color discrimination circuit 256 black area correction section 256a the number of pixels of the same color Counting circuit (calculating means) 256b Dispersion degree calculating circuit (calculating means) 256c Black discriminating circuit (color data replacing means) 257 Second noise erasing section 258 Color replacing section 26 Card drive control section 27 RAM 28 ROM 29 Distance measuring section 30 Photometry Unit 31 Setting color normalization calculation unit 32 Color area setting calculation unit 34 Control unit 35 Whiteboard 36 Character part 37, 38 Misidentified pixel

Claims (2)

[Claims] 1. A color discriminating device for discriminating a color of a color image composed of image data of a plurality of color components, comprising: a color discriminating means for discriminating a color using the image data for each pixel position; Calculating means for extracting data of the color determined by the color determining means in units of a predetermined number of pixels, and calculating a degree of dispersion indicating a degree of dispersion of colors included in the extracted area; A determination unit that determines whether the color variance is large by comparing the degree of dispersion calculated by the unit with a predetermined threshold value set in advance; and when the determination unit determines that the color variance is large, A color discriminating device comprising: color data replacement means for replacing color data at a pixel position at the center of an extraction region with black data. 2. The color discriminating apparatus according to claim 1, wherein said color data replacing means converts color data at a pixel position at the center of said extraction area when said discriminating means determines that the color variance is large. A color discriminating apparatus for replacing color data of the entire extraction area with black data instead.