JPH04354067A - Generating method for antecedent part membership function in fuzzy picture processing - Google Patents

Abstract

PURPOSE:To obtain a color picture close to human sense in the color adjustment of color picture data by a video camera by extracting an area where the color adjustment is executed by the thermal picture data of a subject photographed by an infrared ray camera, and determining a membership function, and executing fuzzy correction. CONSTITUTION:The CCD camera 1 photographs the color picture of a person, and the infrared ray camera 2 photographs the thermal picture of the subject. A color signal RGB from a picture processor 3 is converted into the signal L, Cab, Hab of a color space coordinate system by a converting part 5. A estimating part 6 estimates a flesh color area from the values of L, Cab, Hab. An extracting part 7 extracts the flesh color area by the data of the estimating part 6 and the data from a thermal picture processor 4. A fuzzy correcting part 8 determines the membership function, and gives the color adjustment by fuzzy inference to the signals L, Cab, Hab of the flesh color area. A restoring part 9 restores the color adjusted signals L, Cab, Hab to the RGB signal, and outputs it by combining it with the RGB signal of the other part than the flesh color area.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to color image processing devices such as video input/output devices, and more particularly to a method for creating membership functions for color adjustment using fuzzy inference.

0002

[Prior Art] For example, in a video printer, instead of faithfully outputting an original image inputted by a video camera, the printer produces a color image that is more beautiful than the original image by beautifully reproducing the skin tone of a person, etc. There are times when you want to get it.

[0003] However, it is difficult to understand people's color sense, and it has been difficult to perform color adjustment in consideration of people's sensitivities and senses of beautiful skin color.

[0004] On the other hand, in the 1990 Fuzzy System Symposium Lecture Collection ``Fuzzy Image Processing Considering Sensitivity in Color Copy System'', a selective color adjustment method that takes human sensibilities into account was proposed. ,For example, skin color corresponds to the skin of a person,
There are many types of skin tones, such as natural skin tones such as fruits such as peaches, and artificial skin tones used in building materials, etc., and it is not possible to select only the skin color of a person from these skin tones and perform color adjustment. It was an extremely difficult problem.

[0005] Furthermore, it is extremely complicated and time-consuming to perform color adjustment that takes into account changes in a person's skin color due to changes in lighting, changes in skin color due to white balance imbalance, etc. .

[0006]

Problems to be Solved by the Invention In view of the problems of the prior art, the present invention selects only the skin color of a person's skin from among many skin colors, and performs color adjustment using fuzzy inference to create a human-like image. The purpose is to obtain a color image of a subject that evokes a feeling of being desirable.

[0007]

[Means for Solving the Problems] The present invention provides a fuzzy image processing method in which image data of a subject photographed by a video camera is corrected and color adjusted using a fuzzy inference method using a membership function. , Extracting a region on the image data where color adjustment is to be performed using thermal image data of the subject photographed by an infrared camera, and using color information of the image data within the extracted color adjustment region. The first step is to create the membership function.

[0008]

[Operation] Image data corresponding to the skin tone area of a person's skin is extracted from image data obtained with a video camera using thermal image data of an infrared camera, and color adjustment is applied only to this area using fuzzy inference. In doing so, a membership function in fuzzy inference is determined from color information of image data within the region.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The fuzzy image processing method of the present invention will be explained in detail below with reference to the drawings regarding an embodiment of a video input/output device.

FIG. 1 shows a hardware configuration diagram of a video input/output device. In the figure, 1 is a CCD camera that takes a color image of the subject, 2 is an infrared camera that takes a thermal image of the subject, 3 is an image processing device that processes the color image signal from the CCD camera 1, and 4 is an image processing device that processes the color image signal from the CCD camera 1. a thermal image processing device that processes the thermal image signal from the infrared camera 2; 5 is a conversion unit that converts the color (RGB) signal of the image processing device 3 into an L*Cab*Hab° signal in a color space coordinate system; Reference numeral 6 denotes an estimation unit that estimates the skin color area of the subject from the value of L*Cab*Hab°, and 7 estimates the entire subject using the data of the estimation unit 6 and the thermal image data from the thermal image processing device 4. 8 is L*Ca of the area corresponding to the skin color of the person extracted by the extraction unit 7;
A fuzzy correction unit performs color adjustment on the b*Hab° signal using a membership function based on fuzzy inference, and 9 indicates the color-adjusted L*Cab*Ha obtained by the fuzzy correction unit 8.
An RGB conversion unit 10 restores the b° signal to an RGB signal, and outputs the thus obtained color-adjusted RGB signal and the RGB signal of the part other than the skin color area of the person to a printer to reproduce the image of the subject. This is a printer that performs

FIG. 2 shows one screen obtained by the CCD camera 1, which is composed of n×m images horizontally and vertically. In the figure, i and j indicate arbitrary pixels within this one screen, so they are used to express pixels in a matrix (i, j).

FIG. 3 is a diagram schematically showing an example of a thermal image obtained by the infrared camera 2, and the rightmost band indicates a temperature region. This figure shows that the skin color of a person's skin is seen in the face, neck, and hands, and because the skin in these areas is exposed through clothing, the temperature ranges from 33 to 37 degrees Celsius, which is close to body temperature. . In general, there are a wide variety of skin colors, such as the skin color of skin, natural skin colors such as fruits, and artificial skin colors such as walls, but human skin has a temperature range close to body temperature and can be distinguished from other skin colors. Using the thermal image of the infrared camera 2, it is possible to selectively extract only the skin color.

The operation of the video input/output device having the above configuration will be explained with reference to the operation flowchart of FIG.

First, in step S1, when a start signal is input, the pixels (i, j) of the image processing device 3 and the thermal image processing device 4 are initialized.

[0015]

[Math 1]

is set to .

Next, in step S3, i of pixel (i, j) is compared with n, and if it is smaller, j is compared with m in step S4. If j is smaller than m in step S4, the image data (RGB signal) of the CCD camera 1 is output in step S5, and this is converted to L* in step S6.
Convert to Cab*Hab° and convert the image data (L*Cab*
Hab°) 11 is obtained.

The image data thus obtained (L*Cab
Based on the value of *Hab°) 11, it is determined in step S7 whether or not this is the skin color, and if it is the skin color, matching with the thermal image taken by the infrared camera 2 is performed in the next step S8, and it is determined that this is the skin color of the person's skin. A determination is made whether or not. In this determination, if the image data (L*Cab*Hab°) 11 is a skin color area of a person's skin, step S9
Then, this data is subjected to color adjustment using membership functions based on fuzzy inference. Note that the skin color is determined based on the following Equation 2.

[0019]

[Math 2]

[0020] Here, a method for determining the membership function will be explained. First, read the image data L*Cab*Hab° of the skin color area of the human skin obtained above, and from the read values L*L, L*M, L*S, Cab*L, Ca
b*M, Cab*S, Hab°L, Hab°M, Hab
Determine °S. Here, the subscripts L, M, and S are respectively Large.
It stands for e, Middle, and Small.

Let the value of L*Cab*Hab° just read be the following equation 3.

[0022]

[Math 3]

[0023] However, L*M, Cab*M, H in Equation 3
ab°M was calculated using the following equation 4.

[0024]

[Math 4]

Next, the antecedent part of the membership function is created using the value of L*Cab*Hab° in Equation 3 above. The following is the membership function Small(FHab°
Although the method for creating S) is shown here, L* and Cab* can also be created using the same method.

First, the starting point s, vertex p, and ending point e of FHab°S are determined by the following equation 5.

[0027]

[Math 5]

[0028]FHab°S at this time is expressed by the following equation 6.

[0029]

[Math 6]

When FHab°S obtained using the equation 6 is graphed, it becomes an isosceles triangular graph as shown in FIG.

Similarly, FHab°VS, FHab°M
, FHab°L, FHab°VL,
As shown in FIG. 5, a membership function is obtained in which several isosceles triangles intersect.

Note that FIGS. 6 and 7 similarly show FCab*V
S, FCab*S, FCab*M, FCab*L, FC
ab*VL, FL*VS, FL*S, FL*M, FL
The results obtained for *L and FL*VL are shown in a graph.

By performing color correction by fuzzy inference on the skin color area of a person's skin using each membership function obtained in this way, it is possible to accurately convert the skin color area of the person to a preferable color, and to It becomes possible to reproduce color images that are adapted to the senses.

[0034]

As described above, the present invention provides a fuzzy image processing method for correcting and color adjusting image data of a subject photographed by a video camera using a fuzzy inference method using a membership function. In this step, a region in which color adjustment is to be performed on the image data is extracted using thermal image data of the subject photographed by an infrared camera, and color information of the image data in the extracted color adjustment region is extracted. By creating the membership function from The reproduced image has the effect of reproducing colors that evoke a sense of preference in humans.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an example of the hardware configuration of a video input/output device to which the fuzzy image processing method of the present invention is applied.

FIG. 2 is a plan view showing one screen of a CCD camera.

FIG. 3 is a plan view showing a thermal image of an infrared camera.

FIG. 4 is a flowchart showing the procedure of fuzzy image processing.

FIG. 5 is a graph showing the membership function FHab°.

FIG. 6 is a graph showing the membership function FCab*.

FIG. 7 is a graph showing the membership function FL*.

[Explanation of symbols]

1 CCD camera 2 Infrared camera 3 Image processing device 4 Thermal image processing device

Claims (1)

[Claims] Claim 1. A fuzzy image processing method for correcting and color adjusting image data of a subject photographed by a video camera using a fuzzy inference method using a membership function. Extracting a region to perform color adjustment from thermal image data of the subject photographed by an infrared camera, and determining the membership function from color information of the image data within the extracted color adjustment region. A method for creating an antecedent membership function in fuzzy image processing, characterized in that: