JP3452212B2 - Color image processing device that performs color reduction processing - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color image processing apparatus, and more particularly to a color image processing apparatus for subjecting a full color image to color reduction processing in the field of digital color hard copy such as color printing and color copying.

[0002]

2. Description of the Related Art In the case of digitally forming a color image, a predetermined number of bits are assigned to each color component of R (red), G (green), and B (blue), and the level of each color component is changed. A desired color is reproduced by combining the three primary colors. The types of reproducible colors in this case differ depending on the number of bits representing each color component. That is, as the number of bits increases, the types of colors that can be expressed increase. If you think only by increasing the types of colors that can be expressed, you can increase the number of bits as necessary, but if you increase the number of bits, the burden on hardware and software will increase accordingly, so This causes an increase in cost and an increase in processing time. Therefore, in practice, the number of bits suitable for the device is selected in consideration of the balance between image quality and cost.

For example, when outputting an image to a color monitor such as a personal computer, in a high-end model in which color reproducibility is important, 8 bits are assigned to each color and 1 bit is assigned.
A pixel is represented by 24 bits. On the other hand, one pixel is represented by 8 bits in a popular model that can reproduce colors in a general range.

As described above, since the number of bits for expressing one pixel is different depending on the device, the image data of multi-bit expression used in a certain device may not be directly applied to another device of small bit expression. is there.

Therefore, conventionally, a process of converting image data of multi-bit expression into image data of low-bit expression, that is, a process of reducing the types of reproducible colors has been performed. This processing is called color reduction processing or limited color processing.

Many image processing methods of this type, which are called color reduction processing / limited color processing, have been proposed. These are Step 1: For example, R (red), G (green), B (blue) 8 each
When generating a limited color image of 1 pixel 8 bits from a full color image of 1 bit (24 bits per pixel), first, RGB
A three-dimensional histogram for the three primary colors is generated, and 256 (= 2 ⁸ ) colors having a higher appearance frequency than the histogram are selected and set as the representative color.

Step 2: Next, the representative colors are sorted to determine the color code corresponding to each representative color.

Step 3: After that, a representative color closest to each pixel value of the input image is selected, and a color code corresponding to the representative color is given.

If it is largely divided, the color reduction processing is realized by the above three steps. These processes are very effective. Especially, by using the area gradation expression such as dithering or error diffusion in the step of selecting the representative color, the image quality equivalent to the full color can be obtained with a small number of colors even for a picture image. It is widely known that

For example, in Japanese Patent Laid-Open No. 2-1076,
In the step of selecting the representative color described above, based on the uniform color space, the color distribution is transformed into the principal axis and the histogram projected in the direction of the first principal axis is divided by the threshold value by the discriminant analysis to obtain a good representative image with less image deterioration We are proposing a method to select colors.

Further, in Japanese Patent Laid-Open No. 4-257984, a representative color that does not leak even if it is a color having a low appearance frequency by dividing a color space into a plurality of partial spaces and determining a representative color for each partial space. The selection method is disclosed.

Further, in Japanese Patent Laid-Open No. 2-116893, when the color space is divided in relation to the step of selecting the closest representative color, the parent-child relationship of the division is managed by a tree structure. , Efficient allocation to representative colors is achieved by having a pointer information list for that structure for each space.

[0013]

However, the above-mentioned prior art has the following problems.

First, a large memory capacity is required for the three-dimensional histogram generated when the representative color is selected.

For example, as described above, in order to generate a three-dimensional histogram for an original image having 256 gradations of 8 colors for each RGB color, it is necessary to secure 24 bits≈16,770,000 elements in a memory. However, preparing such a large-capacity memory causes the device to be complicated and the cost to be increased. In order to reduce the memory capacity, it is necessary to reduce the number of gradations (bit number) of each color, but an extreme reduction in the number of gradations deteriorates the quality of the final color-reduced image.

Further, the processing for searching for the closest representative color also becomes large-scale. To obtain a good color-reduced image, RGB
In the three-dimensional color space of, it is desirable to select the representative color with the smallest Euclidean distance from the input pixel value, but for each pixel of the input image, the Euclidean distance from all the representative colors is calculated, and the minimum distance is calculated. A huge amount of calculation is required for the process of searching for a given representative color.

Further, in the prior art, in order to perform rich gradation expression, processing of area gradation expression such as error diffusion and dither is often performed. These processes are not particularly problematic when the target image is a picture image such as a photograph, but there are some problems when targeting images such as business manuscripts including graphs and illustrations and low-priced printing. Dots occur.

In business manuscripts, it is often the case that the number of colors is suppressed and shades of the same series are used in consideration of design and color arrangement. Even in low-priced printing, the number of color plates is reduced from the viewpoint of cost, and as a result, the shades used in one image are limited. Both of them have many elements such as characters and line drawings.

In the area gradation processing, the uniformity of the color in the solid (solid) portion and the gradation (continuous gradation change) portion, which are frequently used in the image, is deteriorated, and the quality in the character and line drawing portions is improved. There is a problem of causing deterioration.

As described above, the conventional color-reduction processing has some problems, and particularly when the target image is a business manuscript, there are problems to be solved in terms of processing scale and image quality.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a color image processing apparatus which causes less deterioration of image quality even when a color reduction process is performed.

[0022]

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems and uses a certain signal in a three-dimensional color space as a reference signal, and determines the ratio of the remaining two signals to the reference signal. /> means for generating a resistance value signal, means for generating a 2-dimensional histogram by measuring the frequency of occurrence of the characteristic value signal, means for calculating a representative characteristic value from the two-dimensional histogram, two characteristic values Means for replacing the signal with any of the representative characteristic values.

Using three primary color signals as the reference signal,
As the characteristic value indicated by the characteristic value signal, the ratio of the remaining two signals to the maximum value signal can be used.

[0024]

For example, by using a system in which the maximum value signal of BGR is used as the reference signal and the ratio of the remaining two signals to the maximum value signal is used as the characteristic value, color reduction processing that limits the ratio of colors such as B: G: R becomes possible. Become. As a result, the histogram to be generated becomes two-dimensional, and the memory capacity can be reduced. Also, since the "color ratio" is limited, it is possible to reduce the number of representative colors themselves without impairing the image quality, and the processed image that limits the "hue" rather than the "color" of the input color image. As a result, a good color-reduced image is generated especially in the color-reduction processing for business manuscripts.

[0025]

EXAMPLES The present invention will be described in detail below with reference to examples.

In this embodiment, B is used as the reference signal.
(Blue) G (green) R (red) maximum value signal, and the ratio of the remaining two signals to the maximum value signal as a characteristic value is used to reduce the color ratio such as B: G: R. I will explain.

FIG. 1 is a diagram showing the form of input / output images of image processing according to this embodiment. FIG. 1A shows an example of an input full-color image, which is a full-color image consisting of 8 bits for each pixel of RGB. FIG. 1B shows a color-reduced image to be output. The output image is represented by 8 bits for each pixel, and the upper 3 bits are a color code representing a hue such as red, blue, and gray. Yes, the lower 5 bits are gradation data indicating the depth of color. One color code and hue (R: G: B ratio) are set for the processed image and added to the image as header information.

As described above, this embodiment converts a full-color image into a reduced-color image composed of a color code and gradation data.

FIG. 2 is a diagram showing the overall construction of an embodiment of a color image color-reducing apparatus according to the present invention. As shown in the figure, the processing here is a typical one which is optimum for the processed image. The process of selecting a color ratio (color code in FIG. 1B) and searching for the closest color ratio to each pixel of the original image,
The process is roughly divided into a process for generating a color-reduced image composed of a color code and gradation data as shown in FIG.

In FIG. 2, the color ratio generation unit 21 uses the BG
A maximum value signal 25 in the R space, a main color signal 26 indicating the maximum value color, and two color ratio signals 27a and 27b indicating the ratio of the remaining two signals to the maximum value signal are generated. The histogram generator 22 generates a two-dimensional histogram 28 from the main color signal 26 and the two color ratio signals 27a and 27b. The representative color ratio extraction unit 23 extracts the representative color ratio from the created two-dimensional histogram. In the subtractive color image generation unit 24, the representative color ratio of gray is assigned in the case of achromatic color, and the representative color ratio table 29 is searched by using the input main color signal and color ratio signal in the case of chromatic color. The closest color ratio is selected and the subtractive color image 210 is generated.

FIG. 3 is a diagram showing a flow of processing for selecting a representative color ratio.

The processing of the color ratio generator 21 will be described. First, the original image is input (step 101), and input RGB
Maximum pixel value Max (B, G, R) and minimum pixel value Min
(B, G, R) is detected (step 102). The signal indicating the maximum value of the input RGB pixel values is the maximum value signal 25. Next, the difference Max-Min between the maximum value and the minimum value is compared with a predetermined threshold value TH (step 103), and Max-Min is compared.
If -Min is larger than TH, the input pixel is regarded as a chromatic color and the processing is continued. If it is smaller than TH, it is judged that the input image is an achromatic color (step 104), and the subsequent processing is not performed. This is because B: G: R = 1: 1: 1 for achromatic color (gray) is set in advance in one of the eight color codes.

Next, the main color signal 26 and the color ratio signal 27
a and 27b are generated. The main color signal 26 is a signal indicating whether the maximum value is the R signal, the G signal, or the B signal (step 105). Also, the color ratio signal 27
a and 27b are 2 other than the main color (maximum value)
It is a signal indicating how large the signal is, and is obtained by the color ratio signal = one signal other than the main color × (255 / maximum value) (steps 106, 107, 108).

The histogram generator 22 is arranged to detect the main color signal 2
It is a means for generating a two-dimensional histogram 28 from the six and two color ratio signals 27a, 27b. Here, since the main color is either RGB, the two-dimensional histogram of the main color R, the two-dimensional histogram of the main color G, and the two of the main color B
Three types of two-dimensional histograms of the three-dimensional histogram are generated. The histogram generation unit 22 selects which of the three histograms to use based on the main color signal 26, and increments the color value represented by the two color ratio signals 27 by one. That is, the address of the two-dimensional histogram defined by the two color ratio signals 27 is incremented.

This processing is performed on the entire image to complete the histogram (steps 109, 110, 11).
1).

The representative color ratio extraction unit 23 is a means for extracting the representative color ratio from the created two-dimensional histogram, and is a known technique of MCA (Median Cut Alg).
(orithm: median division method) is performed on the two-dimensional plane and three surfaces to extract the representative color ratio. FIG. 4 shows the processing content.

FIG. 4 is a diagram showing the outline of a normal MCA algorithm. In the three-dimensional color space, the MCA selects a representative color by dividing the color space by dividing the axis with large variation into two until the number of colors reaches a predetermined number. First,
A variance value or the like is calculated for each RGB axis of each color space, the axis with the largest variation is selected, and then the color space is divided into two by the average value of the axes. In FIG. 4, the axis represented by the thick line indicates the axis with the largest variation. First, in the initial state (1), the variance value of each RGB axis is calculated, it is determined that the R axis with the largest variation is used for division, and the color space is divided by the average value. Two color spaces are obtained. Next, it is decided to use the B axis shown by the thick line with the largest variation among the 6 variance values of the RGB axes of the two color spaces of (2) for division, and the color space is divided by the average value. Then, three color spaces are obtained as shown in (3). Similarly,
The state in which four color spaces are generated by the G-axis division indicated by the thick line is (4). Hereinafter, this division is repeated until the same number of color spaces as the number of representative colors is obtained, and then the average value of the color distribution of each color space is set as the representative color.

FIG. 5 shows a representative color ratio selection process by MCA for the two-dimensional histogram performed in this embodiment. In the present embodiment, since a two-dimensional histogram is generated for each of the main RGB, the processing is started from the initial state of the plane number 3 as shown in (1). First, variance values are calculated for a total of 6 axes, which are 3 plane axes, and the axis having the largest value is used for plane division. Next, the plane is divided by dividing the plane by the average value of the selected axes (steps 112, 113, 114) until it becomes equal to the number of representative color ratios.

In FIG. 5, the thick line indicates the selected axis, and in the initial state of (1), the B axis of the two-dimensional histogram of the main color R has the largest variation, and the average thereof. By the plane division by the value, the state of (2) is obtained. Similarly, the B axis of the main color G and the R axis of the main color G are selected, and the number of planes becomes 6 at the time of (4). In this embodiment,
Since the number of representative color ratios is "8" and one of them is set to achromatic color, another plane division is performed and the number of planes is set to "7". The average value is calculated, a representative color ratio is obtained (step 115), and the representative color ratio table 29 is set. At this time, by clustering the representative color ratios in which of the main colors RGB, the color ratio search in the subsequent subtractive color image generation can be speeded up.

Next, the nearest neighbor color ratio probe processing will be described. FIG. 6 shows the nearest neighbor color ratio search process. The process of generating the color ratio is the same as the process of selecting the representative color ratio, and generates the maximum value signal Max (step 20).
2), chromatic / achromatic color determination (step 203), generation of main color signal and color ratio signal (steps 205-2)
08). Steps 201 to 208 in FIG. 6 correspond to steps 101 to 208 in FIG.

The subtractive color image generator 24 assigns a gray representative color ratio in the case of an achromatic color (step 204), and uses the input main color signal and color ratio signal in the case of a chromatic color to represent the representative color. The ratio table 29 is searched (steps 209, 210, 211) and the closest color ratio is selected. The processing at this time selects the representative color ratio that minimizes the Euclidean distance in the color ratio plane selected by the main colors.
(Step 212). At this time, in the process of limiting the color ratio as in this embodiment, the Euclidean distance calculation is 2
Dimensional distance calculation is sufficient, and the number of representative colors themselves can be reduced to a small number as compared with the conventional limited color processing, so that the amount of calculation in the search processing can be significantly reduced. Next, the maximum value signal M
The upper 5 bits of ax are used as gradation data, and a code corresponding to each representative color ratio is added to the top 3 bits to generate a subtractive color image signal of 8 bits for each pixel (step 213).

By the above processing, a good color-reduced image can be obtained. In the embodiment, the color ratio code is 3 bits and the gradation data is 5 bits, but it goes without saying that the present invention is not particularly limited by these numerical values. Also, BGR
Similar processing can be performed with a subtractive color mixture system of YMC (yellow, magenta, cyan) as well, and in this case, instead of assigning a gray color code in the embodiment, a K (black) signal is used. It is also possible to perform color reduction processing and black plate generation at the same time by generating.

FIG. 7 shows a different embodiment of the invention. This is because the primary image is not a primary color image such as RGB, but L
6 is a conceptual diagram of an example in the case of a luminance / chromaticity separation type color space such as ^* a ^* b ^* . As shown in the figure, a chromaticity plane defined by a certain luminance value is provided, and L ^* a
^* b ^* pixel value and white point (eg L ^* = 100, a ^* =
A two-dimensional histogram is generated with (a ^* , b ^* ) coordinate values of a point where a straight line connecting 0, b ^* = 0) intersects the plane. The (a ^* , b ^* ) coordinates can be easily obtained by scale conversion by linear operation for the input L ^* a ^* b ^* . Next, the representative color point is determined by performing the same plane division as in the above-described embodiment. However, in the present embodiment, since the classification by the main color is unnecessary, the one surface of the two-dimensional histogram is sufficient. Then, (a ^*, b ^*) coordinates by a color code, the distance from the white point ^{(√ (L * 2 + a} * 2 + b * 2) color-reduced image to grayscale data is generated. That is L ^* a
By limiting the straight line extending from the white point in the ^* b ^* space, the same effect as limiting the ratio of R: G: B in the RGB color space is obtained. It can be seen that the same processing as that of the RGB color space can be performed even for the type.

[0044]

As described above, according to the present invention, the three-dimensional color space
Since the characteristic value signal generated as a ratio of the remaining two signals to a certain reference signal is converted into a two-dimensional color plane and the color reduction processing is performed, the memory capacity at the time of histogram generation, the number of representative colors, etc. can be reduced by the conventional limited color processing. It can be greatly reduced compared to. Further, since the reference signal and the characteristic value signal work so as to preserve the tone and gradation of the color in the original image, the quality of so-called business document image including character / line drawing image, solid / gradation image, etc. is impaired. Color reduction processing is possible without any problem.

[Brief description of drawings]

FIG. 1 is a diagram showing a form of an input / output image of image processing according to the present embodiment.

FIG. 2 is a diagram showing an overall configuration of an embodiment of a color image color reduction device according to the present invention.

FIG. 3 is a diagram showing a flow of processing for selecting a representative color ratio.

FIG. 4 is a diagram showing a process of extracting a representative color ratio.

FIG. 5 is a diagram showing a representative color ratio selection process by MCA for a two-dimensional histogram performed in the present embodiment.

FIG. 6 is a diagram showing a nearest neighbor color ratio search process.

FIG. 7 is a diagram showing another embodiment of the present invention.

[Explanation of symbols]

21 ... Color ratio generation unit, 22 ... Histogram generation unit, 23
... Representative color ratio extraction unit, 24 ... Reduced color image generation unit, 29 ... Representative color ratio table

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06T 1/00 510 H04N 1/46 H04N 1/60 G09G 5/02 G09G 5/06 JISST file (JOIS)

Claims (2)

(57) [Claims] 1. A means for generating a ratio of two remaining signals to the reference signal as a characteristic value signal by using a certain signal of a three-dimensional color space as a reference signal, and measuring the appearance frequency of the characteristic value signal. And a means for calculating a representative characteristic value from the two-dimensional histogram, and a means for replacing two characteristic value signals with one of the representative characteristic values. An image processing device that performs processing. Wherein a said reference signal three primary color signals, 2 characteristic values the characteristic value signal indicates the remaining relative maximum value signal
The image processing apparatus according to claim 1, wherein the image ratio is a signal ratio.