JP2623029B2 - Color image processing method - 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 method, and more particularly to a color video printer, a color copying machine, a color display device and the like.

[0002]

2. Description of the Related Art For example, in a color copying machine, there is a case where it is desired to obtain a beautiful color copy image of a flesh color based on an original color image, instead of faithfully copying an input color original image. The applicant has filed Japanese Patent Application No. Hei.
193542. The image processing in this application utilizes fuzzy inference. For example, in order to obtain a beautiful flesh color, a membership function indicating the degree of conformity to the flesh color is defined in advance for a predetermined flesh color area, and fuzzy inference is performed using this membership function to obtain a beautiful flesh color. The amount of color correction to be obtained has been determined.

[0003]

However, in such image processing, a skin color region is determined in advance based on color data, and color correction is performed on the skin color region. differences and when processing an image taken under the colored light source which, when processing an image taken by the video camera was not the white balance can not correct for a desired skin color area and become. Therefore, the present invention is intended to determine a color region to be corrected and a membership function for the region for each screen to be processed.

[0004]

According to the present invention, in a color image screen to be processed, a pixel in a desired area (for example, a flesh color area) to be corrected is designated, and a pixel located around the designated pixel and A region formed by a pixel having a color close to the color of the designated pixel is determined as a correction processing target region, and the actual value of a color attribute (brightness, saturation, hue) of a pixel belonging to the correction target region is determined. On the basis of,
The structure is such that a membership function is determined for each attribute. Then, the membership function is determined based on the maximum value, the intermediate value, and the minimum value of the color attribute values of the pixels belonging to the correction target area.

Further, in a color image screen to be processed, a plurality of pixels in a desired area (for example, a skin color area) to be corrected are designated, and the color attributes (brightness, saturation, hue) of these pixels are specified. The membership function is determined for each attribute based on the actual value of. Then, the membership function is determined based on the maximum value, intermediate value, and minimum value of the color attribute values of these pixels.

[0006]

According to the above arrangement, a correction target area is determined for each color image screen to be processed, and a membership function for the correction target area is determined. a color image screen or not, even for image screen photographed by the video camera was not white balance, it is accurate color correction.

[0007] In particular, a desired region to be corrected (e.g., skin color region of the portion of the face) to specify the pixels in, of the designated pixel
If the region where the pixel is formed having a color close to and to the color of the designated pixel positions in peripheral and configured to be corrected processing target region, the region to be corrected can be specified by the position information such as XY coordinates, The correction is performed only for the skin color of the face part, and no correction is performed for the skin color of the other part.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A case where color image processing is performed so as to obtain a preferred skin color will be described below with reference to the flowchart shown in FIG.

This embodiment is an example of a color video printer. Previously not a, the input color image signal (RGB signal)
Is converted into a color image signal of a color space L (brightness), Cab (saturation) , and Hab (hue) cylindrical coordinate system as shown in FIG. 8 determined by the CIE (International Commission on Illumination). Then the converted L,
It is determined whether or not the Cab and Hab signal data is in the flesh color area. This determination method will be described later. The pixels determined to be in the skin color area are corrected by fuzzy image processing for obtaining a preferable skin color. No fuzzy image processing is performed on the other pixels. The signal subjected to the fuzzy image processing is converted into a YMC color signal suitable for printing, and printing is performed based on this signal.

Hereinafter, the fuzzy image processing will be described. FIGS. 2, 3, and 4 show condition part membership functions for performing fuzzy image processing for L (brightness), Cab (saturation), and Hab (hue), respectively. In this embodiment, the correction color area is set to (vl: sufficiently large,
1: large, m: medium, s: small, vs: sufficiently small) and a membership function is defined for each region.

The membership function is determined as follows for each image to be processed. Here, the processing target image is an image in which a human face is shown. First,
Using the mouse or the like, one point of the skin color portion of the face is designated on the screen projected on the CRT. Then, the color space coordinates of the specified pixel are set to L1 *, a1 *, and b1 * (see FIG. 8), and the color space coordinates of the adjacent pixel are set to L2.
When *, a2 *, and b2 * are set, it is determined whether or not the following expression is satisfied, and if satisfied, the area is expanded in such a manner that adjacent pixels are further determined to form a pixel that satisfies the following expression The area is set as a correction target area.

[0012]

[Number 1] ^{(L1 * -L2 *) 2 +} (a1 * -a2 *) 2 + (b1 * -b2 *) 2 <k

In this equation, k is a threshold constant, which is set to 10 in the embodiment. However, if the user can arbitrarily specify, a more versatile image correction process can be performed. Further, when the correction target area is determined, RGB values may be used without using values of lightness, saturation, and hue of a color.

In the above method, the correction target area is specified as the pixel position information. However, a plurality of portions of the skin color portion of the face are displayed on the screen projected on the CRT using a mouse or the like. Point out, lightness L, saturation Cab, hue H
The value of ab is obtained, and then the lightness L, the saturation Cab, and the hue Hab
, The maximum value and the minimum value may be obtained, and the correction target area may be determined based on these values. In this case, the value of the designated point may not be adopted as the value of the designated point, and the average value of the surrounding values may be used as the value of the designated point.

The condition part membership function of the fuzzy image processing rule is determined for the color of the correction target area as follows.

First, a maximum value, a minimum value, and an intermediate value are obtained for the lightness L, the saturation Cab, and the hue Hab of the pixel belonging to the correction target area.

For a face portion of a screen on which a human face is photographed, lightness L, saturation Cab, hue Ha are designated by specifying a plurality of locations.
In the case where the correction target area is determined by obtaining b, when the value is actually obtained, Llarge = 64, Lmidd
le = 37, Lsmall = 10, Cablage =
25, Cabmiddle = 20, Cabsmall =
15, Hablarge = 126, Habmiddle
= 90, Habsmall = 54. In this embodiment, the intermediate value is not an actual measured value but an average value of the maximum value and the minimum value.

The membership function is determined based on these values, 2, 3, is shown in Figure 4.

For example, the hue Hab will be described.
The membership function Habs has a starting point of Habsmal.
1- (Habmiddle-Habsmall) (1
8) The vertices are Habsmall (54), the end point is Habsmall
The triangle becomes middle (90). Similarly, the membership function Habm sets the starting point to Habsm.
all (54), vertices are Habmiddle (90),
It becomes a triangle whose end point is Hablage (126). The membership function Habl defines the starting point as Habmi.
ddle (90), and the vertex is defined by Habrage (12
6), the end point was defined as Habage + (Habrage-
(Habmiddle) (162).

The membership function Havs is defined as Hab
Is less than 18, the membership value is 1 and Ha
When b is 18 to 54, the trapezoid has a membership value of 1 to 0. Membership function H
abvl is a trapezoid in which the membership value takes a value from 0 to 1 when Hab is 126 to 162, and becomes 1 when Hab is larger than 162. These trapezoidal membership functions are membership functions indicating the extent to which they do not belong to the correction target area.
Such a membership function is necessary to prevent the color correction amount from becoming discontinuous at the boundary of the correction target area.

Similarly, the membership function can be determined for the lightness L and the saturation Cab.

FIGS. 5, 6, and 7 show rules for performing fuzzy image processing. In this example, for L (brightness), Cab (saturation), and Hab (hue), the amount of correction for obtaining a preferable skin color for each state of L, Cab, and Hab is set for each of L, Cab, and Hab. It is determined. Such a rule is determined by obtaining an optimal correction amount after several trials and errors.

L, Cab, Hab for obtaining preferred skin color
The final correction amounts ΔL, ΔCab, and ΔHab are as shown in Expressions 2, 3, and 4, respectively.

[0024]

ΔL = Σ (μi × ΔLi) / Σμi

[0025]

ΔCab = Σ (μi × ΔCabi) / Σμi

[0026]

ΔHab = Σ (μi × ΔHabi) / Σμi

Here, μi indicates the smallest value among the membership values of L, Cab, and Hab for the i-th rule. ΔLi is the correction amount of L by the i-th rule, ΔCabi is the correction amount of Cab by the i-th rule, and ΔHabi is the correction amount of Hab by the i-th rule.

As can be understood from the above equation, for example, L
When the final correction amount ΔL is obtained, L is large, Cab is large, and H is large.
As the degree μi of adopting the correction amount +3 during ab,
The minimum value among the membership values indicating the degree to which each of L, Cab, and Hab belongs to this area is adopted. The final correction amount is determined by the weighted barycenter method.

In this example, the consequent part of the fuzzy rule is:
Although the correction amount is determined as a constant, this is to increase the processing speed. Of course, the consequent part may be defined by a membership function. By adding the final correction amount thus obtained to the original signal, preferable skin color correction for the original signal is performed. That is, L = L +
Calculate ΔL, Cab = Cab + ΔCab, Hab = Hab + ΔHab.

The above is the correction for one data (pixel) of the image signal. This correction is performed for all pixels in the skin color area. The signals for color printing are obtained by converting the above signals into Y (yellow), M (magenta), and C (cyan) signals and outputting them. When the image is displayed on a CRT display device, it is converted into an RGB signal.

In the above-described embodiment, the image processing is performed so as to obtain a preferable flesh color. However, the present invention is not limited to this. For example, a preferable blue sky color or a beautiful leaf color is obtained. It can be applied to such image processing. Of course, in this case, a fuzzy image processing rule corresponding to each case is set.

[0032]

According to the present invention described above, for example, when performing image correction processing for obtaining a desirable skin color, a membership function can be set for each image to be corrected. Therefore, accurate color correction can be performed even on a color image screen that has not been shot under a standard light source or an image screen that has been shot with a video camera having a white balance.

In particular, a pixel in a desired area to be corrected (for example, a skin color area of a face) is designated, and a pixel located near the designated pixel and having a color close to the designated pixel is formed. When the region is set as the correction processing target region, the correction target region can be specified by position information such as XY coordinates, so that only the skin color of the face portion is corrected, and there is no need to correct. There is no inconvenience that the correction is performed for the skin color of the other part.

[Brief description of the drawings]

FIG. 1 is a flowchart showing an image processing method according to the present invention.

FIG. 2 is a diagram showing a condition part membership function of lightness L;

FIG. 3 is a diagram showing a condition part membership function of the saturation Cab.

FIG. 4 is a diagram showing a condition part membership function of a hue Hab.

FIG. 5 is a diagram showing a fuzzy rule relating to brightness.

FIG. 6 is a diagram showing a fuzzy rule regarding saturation.

FIG. 7 is a diagram showing a fuzzy rule relating to hue.

FIG. 8 is a diagram showing a lightness, a saturation, and a hue cylindrical coordinate system color space;

[Explanation of symbols]

 Lvs is a membership function with sufficiently low brightness. Ls is a membership function with low brightness. Lm is a membership function with medium brightness. Ll is a membership function with high brightness. Lvl is a membership function with sufficiently high brightness. Sufficiently small membership function Cabs is a low-saturation membership function. Cabm is a medium-saturation membership function. Cabl is a high-saturation membership function. Cabvl is a sufficiently high-saturation membership function. Is a small hue membership function. Habs is a small hue membership function. Habm is a membership function in hue. Habl is a large hue membership function. Habvl is a sufficiently large hue membership function.

Claims (4)

(57) [Claims] 1. A method for determining a membership value of an input color image signal for an area to be corrected based on a membership function defined for the area to be corrected, and based on the membership value. A color image processing method in which a correction amount for the color image signal is determined, and the color image signal is corrected based on the correction amount, and a pixel in a desired area to be corrected on the color image screen is specified. A region formed around a specified pixel and having a color close to the color of the specified pixel is determined as the correction target region, and the actual value of the color attribute of the pixel belonging to the correction target region is determined. A color image processing method, wherein the membership function is determined based on 2. The color image processing according to claim 1, wherein the membership function is determined based on the maximum value, the intermediate value, and the minimum value of the color attribute values of the pixels belonging to the correction target area. Method. 3. A membership value of an input color image signal for the correction target area is determined based on a membership function defined for the correction processing target area, and based on the membership value. A color image processing method configured to determine a correction amount for the color image signal and correct the color image signal based on the correction amount, wherein a plurality of pixels in a desired area to be corrected on a color image screen are designated. And determining the membership function based on actual values of color attributes of the pixels. 4. The color image processing according to claim 3, wherein the membership function is determined based on a maximum value, an intermediate value, and a minimum value of the color attribute values of the plurality of designated pixels. Method.