JPH03230281A - Area dividing device for color image - Google Patents

Abstract

PURPOSE:To rapidly divide a color image into areas having uniform characteristics by dividing a video signal into plural areas corresponding to color intensity to be the maximum value of differences among respective components of the video signal and then dividing each divided area into plural areas corresponding to its characteristics. CONSTITUTION:An image inputted to an input part 1 is supplied to a color area dividing part 2 and divided into high color, low color and colorless areas. A high color area dividing part 3 divides an area having a prescribed color information out of the high color area divided by the dividing part 2, a low color area dividing part 4 divides the low color area divided by the part 2 furthermore based upon color information and a colorless area dividing part 5 divides the colorless area divided by the part 2 furthermore based upon the color information. Consequently, an area capable of utilizing a feature included in each area can quickly be extracted.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] This invention relates to a color image region segmentation device, and in particular, when measuring or recognizing an object in a video, color information is extracted from an input color image. The present invention relates to a color image region segmentation device that extracts a target by dividing the color image into regions whose characteristics are similar to -.

[Prior art and problems to be solved by the invention] As one means of realizing communication with a sense of presence, images of multiple people in remote locations are synthesized in a virtually created conference space, and
A virtual space communication conference system for communication is being considered. In such a system, it is necessary to extract only human images from images input by a camera or the like.

In order to extract such a human image, a method of dividing a color image of a human image into regions is used.

The conventional method converts the color components of the input video signal, that is, the R (red) component, G (green) component, and B (blue) component, into a feature space, and then performs threshold processing on the histogram within that space. There is a method of performing area division. However, a problem with such a method is that errors are added during conversion to the feature space, and noise that does not exist in reality appears, resulting in erroneous extraction.

On the other hand, instead of a single feature space, there are various feature spaces (R,
G, B, Y, 1. Q, H: Hue, v
: brightness, S: saturation), and recursively divide the area using a histogram of a representative feature space. Since this method simply selects and divides a representative feature space based on the shape of the histogram, it has the drawbacks that unnecessary divisions or too fine divisions are performed, and that processing takes a long time.

Therefore, the main object of the present invention is to provide a color image region segmentation device that can eliminate the above-mentioned drawbacks and can quickly segment a color image into regions with similar characteristics. .

[Means for Solving the Problems] The present invention is a color image region dividing device that divides a color image whose color intensity differs depending on the region into respective regions, and divides the color image into each component of a video signal. a first region dividing means that divides the divided regions into a plurality of regions according to the intensity of the color that is the maximum value difference between the two; and further divides each divided region into a plurality of regions according to its characteristics. and second area dividing means.

[Operation] The color image region dividing device according to the present invention divides a video signal into a plurality of regions according to the color intensity, which is the maximum value of the difference between each component, and further divides each divided region into its own. Since it is further divided into multiple areas according to the characteristics, the histogram of color components in areas with strong colors,
Regions corresponding to all peaks, including small ones, can be extracted.

[Embodiments of the invention] FIG. 1 is a schematic block diagram of an embodiment of the present invention.

First, the configuration of an embodiment of the present invention will be described with reference to FIG. The image input to the input section 1 is supplied to the color region dividing section 2, and is divided into high color, low color, and achromatic regions. The high color area dividing unit 3 divides the high color areas divided by the color area dividing unit 2 into areas having predetermined color information, and the low color area dividing unit 4 divides the high color areas divided by the color area dividing unit 2 into areas having predetermined color information. The color area is further divided using the color information, and the colorless area dividing section 5 further divides the colorless area divided by the color area dividing section 2 using the color information. High color area dividing section 3. The respective regions divided by the low color region dividing section 4 and the colorless region dividing section 5 are integrated by the output section 6 and output as label data representing the divided regions.

Figure 2 is a diagram showing an example of a mask matrix showing high color, low color, and colorless areas, Figure 3 is a diagram showing a method for extracting areas, and Figure 4 is a diagram showing area division of high color areas. Figure 5 is a diagram showing the results of IJ socks; Figure 5 is a diagram showing the matrix representing the results of region segmentation for low color areas; Figure 6 is a diagram showing the matrix representing the results of region segmentation for achromatic areas. FIG. 7 is a diagram showing a matrix showing the integrated division results.

Next, with reference to FIGS. 1 to 7, specific operations of an embodiment of the present invention will be described.

Input unit 1 receives three matrices Rhv, Ghv, and BhV each consisting of a horizontal H image and a vertical V image (H and V are positive integers) obtained by digitizing a color image. These correspond to the red, green, and blue signals of color video, respectively. The color area dividing unit 2 calculates each image of the input image using the following arithmetic formula, and calculates a numerical value "2" representing a high color area, a value "1" representing a low color area, and a value representing a colorless area. A mask matrix Mhv as shown in FIG. 2 is created which has a value of "0" and has the same size vertically and horizontally as the input image.

By processing regions corresponding to each of the values of this matrix using different techniques, it is possible to perform accurate region extraction that was not possible with conventional methods.

Mhv=O: Dhv≦D1 1 :T1<Dhv<Th 2:Th≦Dhv Dhv=MAX (Rhv, Ghv, Bhv)-MIN
(Rhv, Ghv, Bhv) MAX (Rhv, Ghv, Bhv): Maximum value of each component in a certain pixel v of the input image matrix.

MIN (Rhv, Ghv, Bhv): Minimum value of each component at a certain pixel v of the input image matrix.

Here, TI and Th are parameters representing the intensity of color, and T1 is a value that divides the colorless portion and the low color portion. Th is a value that divides the low color portion and the high color portion.

0 or a very small value is given as the value of T1. For example, if one pixel is digitized with 8 bits, T1 is set to 2 or 3. For example, Th may have a value of 43. This is because if the intensity of color is greater than this value, it is guaranteed that there are 256 or more hues at any lightness. These values may be predefined in the device, or may be input in a format during execution.

The high color area dividing unit 3 selects a mask matrix Mhv generated by the color area dividing unit 2 whose value is “2”.
Processing is performed only on the pixel area that is . In this area, the area is further divided using color information. As the color information, for example, hue can be used. rcolo from video signals Rhv, Ghv, Bhv
r garmut transform pai
rs (C.

Mputer Graphics 12.3 A
The hue matrix Hhv is determined using a conversion formula such as that described in "Ug, 1978, pp. 12-19)". This hue matrix Hhv is expressed as an integer value from "0" to r255J for each pixel. In the hue matrix Hhv, create a table that calculates the number of occurrences of that hue value, and calculate the peak of the curve created from the number of occurrences.
Determine the threshold value from the valley as shown in Figure 3, divide the area,
r21J and r22J corresponding to the divided areas.

A matrix as shown in FIG. 4 with labels such as "23" is output.

The low color area dividing unit 4 processes only the areas of pixels whose value is “1” in the mask matrix MhV generated by the color area dividing unit 2. Here, the data of the video signals Rhv, Ghv, and Bhv are used as they are to divide the area, and rl IJ corresponding to the divided area.

A matrix as shown in FIG. 5 with labels such as "12" is output.

The achromatic region dividing section 5 processes only the region of pixels whose value is rOJ out of the mask matrix MhV generated by the color region dividing section 2. Video signals Rhv, Ghv.

A brightness matrix vhV is determined from Bhv using a conversion formula. This brightness matrix Vhv is expressed as an integer value from "0" to r255J for each pixel. Create a table using the number of occurrences of the brightness value in the brightness matrix Vhv, and divide the region by determining the threshold value from the peaks and valleys of the curve created from the number of occurrences in the same manner as in the color region dividing section 3, rolJ, 6th with a label such as r02J
Outputs a matrix like the one shown in the figure.

In the output section 6, each region dividing section 3. 4. In each matrix created in step 5, create one matrix as shown in Figure 7 by integrating all the pixels whose values are not 0, and use this as label data representing the divided areas. It costs 8.

[Effects of the Invention] As described above, according to the present invention, by first dividing an input video into three regions: high color, low color, and achromatic, region extraction can be performed that takes advantage of the characteristics of each region. can be done. Moreover, since the processing is relatively simple and there is no need to perform recursive region division used in conventional methods, the processing can be performed at a considerably high speed.

[Brief explanation of drawings]

FIG. 1 is a schematic block diagram of an embodiment of the present invention. FIG. 2 is a diagram showing an example of a mask matrix showing high color, low color, and achromatic areas. FIG. 3 is a diagram showing a region extraction method. FIG. 4 is a diagram showing a matrix representing the result of region segmentation of a high color region. FIG. 5 is a diagram showing a matrix representing the result of region segmentation of a low color region. FIG. 6 is a diagram showing a matrix representing the result of region segmentation of a high color region. FIG. 7 is a diagram showing a matrix showing the integrated division results. In the figure, 1 is an input section, 2 is a color region dividing section, 3 is a high color region dividing section, 4 is a low color region dividing section, 5 is a colorless region dividing section, and 6 is an output section.

Claims (1)

[Claims] A color image region dividing device that divides a color image having different color intensities into respective regions, the color image being divided by the maximum value of the difference between each component of a video signal. A first region dividing means that divides each region into a plurality of regions according to the strength of a certain color, and each region divided by the first region dividing means,
A color image region dividing device comprising a second region dividing means that further divides the color image into a plurality of regions according to the characteristics thereof.