JPS60193077A - Sorting-out system of color graph - Google Patents

Abstract

PURPOSE:To lighten the burden imposed on hardware by inputting a color graph as black and white various density data to a computer and by extracting the specific color area only by utilizing the black and white various density data. CONSTITUTION:A color graph is drawn with selected colors on a color document 10, which is inputted to a picture input device 11 and read out as black and white various density data. A processor 13 uses a RAM15 as a buffer of the black/white various density data or a register of an intermediate result, etc., with respect to the black and white various density data read out by the picture input device 11 and extracts a color graph corresponding to the specific color. The graph with the specific color extracted by the processor 13 is displayed in the black-and -white or a pseudo color by a picture output device 16.

Description

[Detailed Description of the Invention] [Technical Field] The present invention corresponds to one or several arbitrary colors from so-called color graphs in which system diagrams, route diagrams, wiring diagrams, etc. are drawn using a plurality of colors. This paper relates to a method for selecting graphs to be used.

[Prior art]

A color document image excluding color photographs and text is called a color graph. Color graphs can be processed without using complex algorithms for color photographs, color slides, etc. In other words, from a coloristic standpoint, to classify all colors into categories, three sets of ↑ and 〒 signs such as (R・G−B), (X−Y, −Z) (7) is necessary. Here, l is red,
G means green, B means blue, X means brightness, Y means saturation, and Z means hue. However, when targeting color document images, in order to simplify processing, only Munsell's hue, which has a strong perceptual element, is selected as a feature quantity, and by using the x-y distribution, which has a distribution close to the color wheel, It is possible to realize efficient region segmentation processing based on power category classification for -dimensional feature quantities that is compatible with perceived colors.

However, the method using only Munsell's hue lacks the concept of brightness and saturation, which are perceptually important just like hue. It's impossible. Furthermore, in order to determine the x-y distribution, it is necessary to process a large amount of image data divided into R-G-B data at the image input device stage, which places a very heavy burden on the hardware.

[Purpose of the invention]

An object of the present invention is to provide a color graph sorting method that makes it possible to easily extract graphs corresponding to arbitrary colors on a color graph without placing too much burden on hardware.

[Summary of the invention]

When examining the relationship between black and white gradation data, which can be obtained more easily than RGB data, and color information, it is found that there is a strong correlation between the gradation values of the data and Munsell lightness. Therefore, in the present invention, colors with different lightness are specified in advance, a color graph is drawn using the colors, and the color graph with the specified color is input into a computer as black and white shading data, and only this black and white shading data is used. Extract a specific color area using .

The processing algorithm uses the average gray value of the run for runs between slope points that represent the characteristics of color change points in each block (each scan line) of a black and white gray image, and calculates whether the run is part of a specific color area. We will decide whether it is. Since this process only requires a buffer for several blocks for black-and-white grayscale data, it is easy to implement it in hardware, and there are many possibilities for realizing real-time processing. In addition, by selecting colors based on Munsell's lightness, it is actually possible to divide and extract a number of color regions, so this method has sufficient capability for processing color graphs.

[Embodiments of the invention]

EMBODIMENT OF THE INVENTION Below, one embodiment of the present invention will be described in detail with reference to the drawings.

First, we will explain how to specify colors in a color graph. Among the three attributes of Munsell's color display, brightness has a deep correlation with the relative reflectance of light in the visual sense. Therefore, if a color image input device is specified, it can be considered that the gradation value and brightness have an appropriate correlation. Figure 1 shows the shading values of a total of 18 colors selected from Jacol 220 (a collection of color charts with Munsell notation) from 1.0 to 9.5 in brightness increments of 0.5. , its average gradation value, and standard deviation are summarized. Further, FIG. 2 is a graph showing the relationship between brightness and shading values for the 18 colors shown in FIG. 1.

However, brightness and shading values are not completely proportional.

It can be said that the color with a larger lightness has a larger average gradation value. FIG. 3 similarly shows the relationship between the square root of the average density value and the brightness for the 18 colors shown in FIG. 1, and the relationship is approximately proportional.

Next, in order to know the relationship between saturation, hue, and shading value among Munsell's three attributes, we selected several colors with lightness of 9.0 and 7° 5.4.0, and examined their shading values. I tried it.

Figure 4 summarizes the selected colors, average shading values, and standard deviations. From this, it can be seen that if the brightness is the same, there is no extreme change in the average gradation value due to differences in hue and saturation. However, there are the following two trends.

■The shading value of a color close to blue (with a short wavelength) is smaller than that of a color close to yellow (with a long wavelength). This seems to be due to the strong influence of the spectral characteristics of the optical system such as a drum scanner used in the image input device.

2. For colors with the same hue and brightness, the greater the saturation value, the greater the shade value.

The relationship between tritropism and gradation values in the Munsell color display has been described above, and it can be seen from these facts that the colors constituting the color graph can be specified in a linear order.

(1) Select colors so that they all have different lightness and are not biased in terms of other attributes.

(11) Measure the shading value of the selected color, consider the standard deviation from the average shading value, and leave only colors that do not overlap each other.

In other words, after making a rough selection in step (1),
), the colors should be selected more precisely, taking into account the variation in the gradation values. For example, the 18 colors selected in Figure 1 all have different lightness. Therefore, in the graph of Figure 2, if you select colors so that the average gray value for each color and the line segment with the width of the standard Qli deviation in the center do not overlap, then the graph in Figure 1
It will look like the 12 colors marked with *. It should be noted that the groups indicated by ■ in FIG. 1 are color groups that could not be divided in actual processing.

Next, a process for extracting a specific color area from a color graph, which is the central feature of the present invention, will be explained.

FIG. 5 shows the overall hardware configuration. In the figure, a color graph is drawn on a color document 10 using the colors selected by the method described above.

The document 10 on which this color graph is drawn is transferred to the image input device 1
1 and read it as black and white grayscale data. As the image input bag all, for example, a drum scanner, a book scanner, etc. using a solid-state image sensor is used. Processing device 1
3, the RAM 15 is used as a buffer for the black and white gray data read by the image input device 11 or as a register for intermediate results to extract a color graph corresponding to a specific color. Parameters required for the extraction process in the processing device 13 are input using the keyboard 12. In addition, the necessary programs are in ROM l/
It is provided in l. The graph of the specific color extracted by the processing device [13] is displayed as black and white or pseudo color by the image output device 16.

Image output device 16 is a printer or a display. Note that the extracted graph data may be transmitted to a remote similar device through a communication line.

FIG. 6 shows a flowchart of the algorithm for extracting a specific color area from a color graph. This algorithm uses a constant color run (hereinafter simply referred to as color run) between slope points that indicate the characteristics of color change points in l blocks (l scan lines).
Based on this idea, processing proceeds while always referring to the positional relationship of the upper and lower blocks with the color run.

It also has the following features:

10 Masking of Document (Texture) Part 2, Noise Removal and Correction Using Positional Shift The processing of each step in the flowchart of FIG. 6 will be described below.

Step 1 The following five parameters are input in step l.

LAP: Threshold value of the average Laplacian operator value of the run (for masking text parts) DIST: Threshold value of the positional deviation of the run (for noise removal and correction) MAXGL, MINGL: Upper and lower limits of the average gray value of the run (For specifying a specific color) S P T L-1 Nithrop point I ~ Threshold value for detection To explain the masking of the text part here, the text part changes finely in gradation value, so the response of the light receiving tube is Due to the influence of the characteristics, the original color shading value will not be achieved. on the other hand,
When we try the Laplacian operator on text parts, the value is higher than on non-text parts. Therefore, by averaging the values of the Laplacian operator in runs in the character part and setting an appropriate threshold value (LAP), masking of the character part becomes possible.

Step 2 In Step 2, blocks of black and white gray data (scanning lines)
are input one after another and smoothed with a 3x3 mesh, and then a slope point indicating the characteristics of the color change point is detected. Figure 7 shows the relationship between the original image data (black and white? r!, light color data, smoothed data, and slope point determination vector. Figure 7 (a) shows the original image data for 5 blocks (5 scanning lines). Of these, 4 blocks are reserved in the memory (RAM 15) sequentially for calculation of the Laplacian operator.Every time one block of original image data is input, the grayscale values of the 3x3 mesh are Apply smoothing to take the average of.

FIG. 7(b) shows smoothed data. This smoothed data is compared with a threshold value 5PTH to detect a slope point. FIG. 7(c) is a slope point determination vector, and if the corresponding pixel is a slope point, rt 111. Otherwise, take the value II OII.

This 1107+ run becomes a color run candidate.

Note that slope point 1- is defined as follows.

Assuming that there are no two colors with exactly the same average gradation value in the target color Bunshi image, there will be a characteristic change in gradation value near the boundary of each color area. For this reason, after smoothing with 3X3 mesh,
When examining changes in the grayscale values of pixels in a certain l block (l scanning line), it is found that the grayscale values rapidly increase or decrease at the area where the color changes at the boundary of the area. Therefore, an appropriate threshold value (SPTH) is set to detect a point where the gray value suddenly changes, and that pixel is defined as a slope point indicating the characteristic of the color change point.

Step 3 In Step 3, for each color run candidate obtained in Step 2, the average gray value, average Laplacian operator value, starting point, and run length are calculated. Then, among the color run candidates, only those whose average Laplacian operator value is smaller than LAP and whose average gray value is between MAXGL and MINGL are subjected to the next step 4.

Step 4 In step 4, if at least one of the following conditions is satisfied for three consecutive blocks (three scanning lines), the color run is output as a part of the designated color area. Here, the consecutive blocks are a from the top.

Let them be b and c.

When there are color runs in 1 and b, and there is a color run in a or C whose start point coordinate shift and end point coordinate shift are each smaller than DIST, the color run of b is output.

2. There are color runs in a and C in which the deviation of the starting point coordinates and the deviation of the ending point coordinates are each smaller than DIST.

If b does not have the same color run, create and output the same color run as a in b.

FIG. 8 shows an example of conditions 1 and 2, and (
Figures a) and (b) are examples of condition 1, and figure (c) is an example of condition 2. That is, in step 4, the relationship between the color runs in the three blocks is used, color runs that are extremely far apart are removed as noise, and only one color run in consecutive blocks is removed. When there is a break, correction processing is performed.

Step 5 In step 5, a binary image is output in which only the designated color area having a gray level value between 1MAXGL and MINGL is black, and the rest is white.

-Step 6 In Step 6, it is determined whether processing has been completed for all blocks of the original image data, and it must be completed.]
If t, repeat the process from step 2 onwards.

When selecting specific constituent colors of a color graph using the above-described method, the range of shading values of each color is considered as (average)±(standard deviation). If we assume that the distribution of grayscale values is a normal distribution for each color, then in that range, 68 out of the number of pixels of each color
．． 3%, and pixels with unstable gray values are not included. However, the reason why the specific color region extraction method according to the present invention shows good processing results is as follows.

1. Since it is determined whether the specified color is a specified color using the average gray value for each color run, the influence of pixels with abnormal gray values is small.

2.3 The positional relationship of runs is examined in consecutive blocks (scanning lines) of the tree, and breaks are corrected and noise is removed.

〔Effect of the invention〕

According to the present invention, it is possible to extract a graph of any color from a color graph created using a large number of colors by handling only black and white shading data on a computer.
It is not necessary to process a large amount of color image data, and the burden on hardware can be reduced. Moreover, the extraction accuracy of specific color regions is high, and if used in a system that uses, for example, a color inkjet printer as a color graph output device, a fixed image sensor as an image input device, and a book scanner, etc., the effect will be extremely effective. big. In other words, in such a system, the output image has color information again, making it easier to understand perceptually.
Processing is easy on a computer because it only deals with black-and-white grayscale data. Therefore, it is extremely effective as a system for preprocessing applications such as information compression, storage, and transmission.

[Brief explanation of the drawing]

Figure 1 shows the average shading value and standard deviation of 18 selected colors, Figure 2 shows the relationship between lightness and shading value for the 18 colors in Figure 1, and Figure 3 shows the relationship between lightness and shading value for the 18 colors in Figure 1. Figure 4 shows the relationship between brightness and the square root of the average gray value for the 18 colors in Figure 1.
The figure shows the shading values of colors with the same brightness, Figure 5 is a schematic block diagram of an example of a hardware configuration for realizing the method of the present invention, and Figure 6 shows an example of the process flow for extracting a specific color area according to the present invention. FIG. 7 is a diagram showing the relationship between original image data, smoothed data, and slope point determination vectors, and FIG. 8 is a diagram showing an example of correction due to positional deviation. 10... Color document, 11... Image input device. 12... Keyboard, 13... Processing device, 14...
- ROM, 15...RAM, 16...image output device. Figure 1 (C) (0,1,0,0,0,1,--9)(α)

Claims (2)

[Claims] (1) Read a color graph drawn using multiple colors as black-and-white grayscale data, detect slope points that indicate the characteristics of color change points in the black-and-white grayscale data, and detect the grayscale values of runs between the slope points. A color graph sorting method characterized by determining whether a corresponding run is a specific color area or not, and extracting a graph corresponding to an arbitrary color in the color graph. (2) In creating the color graph, the three attributes of the corresponding color (
Claim 1, characterized in that a set of specific colors determined based on (brightness, saturation, hue) is specified in advance and created.
Color graph sorting method described in section.