JP2001331795A - Device and method for normalizing image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate a face image in a fixed condition in which the variation of contrast by the light and darkness of the image is suppressed by emphasizing the contrast of the dark part of an image and suppressing the contrast of the light part of the image on the other hand. SOLUTION: This device has a shading removing part 14 for obtaining the shading quantity of the face image, which is extracted from an input image and has its size and position normalized, a contrast correcting part 15 for increasing and reducing a luminance value based on the shading value, a distribution correcting part 16 converting the luminance value so that the average and distribution of luminance become previously set values and an average correcting part 17 to normalize contrast.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is suitable for, for example, a security system for authenticating an individual using a face image or a man-machine interface for transmitting human intentions and emotions to a computer from facial expressions and movements. Image normalization apparatus and method.

[0002]

2. Description of the Related Art Hitherto, research has been conducted on a security system for authenticating an individual using a face image and a man-machine interface for transmitting human intentions and emotions to a computer from facial expressions and movements.

[0003] In particular, in order to make the man-machine interface between a computer and a person more natural, the communication between the persons is taken as an example, and the computer is used to identify the individual from the face image, and to express the facial expressions and gestures. Research is being conducted to make people recognize and estimate human intentions and emotions.

[0004] By the way, when authentication using a face image is performed in a security system, or when a computer is used to specify an individual from a face image in a man-machine interface between a computer and a person, the face image is not identified. Absorb changes in size, position, lighting, etc.
It is necessary to always use a face image under a certain condition.

[0005] As a conventional technique for generating a face image under substantially constant conditions, for example, a document “HA Rowley, S. Bal.
uja, and T.Kanada "Human Face Detection in Visual S
For the scenes "CMU-CS-95-158R, Nov. 1995 p3", first remove the background of the face image by using a face image in which the positions of both eyes or both eyes and the mouth are set in advance. For this purpose, the peripheral area of the face image is masked, the shading of the brightness of the face image is corrected, the histogram of the brightness of the face image is obtained, and the histogram is converted so as to be uniform. Such a technique is disclosed.

[0006]

However, generally, in an image, the contrast of a relatively bright area is large, and the contrast of a dark area is small. Also,
In the histogram equalization, the contrast around a luminance value having a relatively large number of pixels tends to be emphasized, so that the contrast in a considerably dark area cannot be increased. Therefore, it is difficult to always generate a face image under a certain condition by the above-described conventional technology.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described problems, and enhances the contrast of a dark portion of an image while suppressing the contrast of a bright portion of the image. It is an object of the present invention to provide an image normalizing apparatus and method capable of generating an image under constant conditions in which a change is suppressed (contrast is normalized).

[0008]

According to a first aspect of the present invention, there is provided an image normalizing apparatus for solving the above-mentioned problems, comprising: a shading calculating means for obtaining a shading amount of an image luminance; There is an increasing / decreasing means for increasing / decreasing the luminance value based on the shading amount, and a converting means for converting the luminance value from the increasing / decreasing means so that the average value and the variance of the luminance become predetermined values.

According to a second aspect of the present invention, there is provided an image normalizing method, comprising: a first step of obtaining a shading amount of an image luminance;
A second step of increasing or decreasing the luminance value based on the shading amount, and a second step of converting the luminance value formed in the second step so that the average value and the variance of the luminance become predetermined values. 3 steps.

That is, according to the image normalizing apparatus and method of the present invention, the contrast of the dark part of the image is enhanced, while the contrast of the bright part of the image is suppressed, and the contrast based on the brightness of the image is normalized. .

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows a schematic configuration of an image normalizing apparatus as an embodiment of the image normalizing apparatus and method according to the present invention.

With reference to FIG. 1, an overall configuration of the image normalizing apparatus according to the present embodiment will be described.

The image input unit 10 receives an input image 20 as shown in FIG. 2 including a face image taken by a video camera or the like. The signal of the input image 20 captured by the image input unit 10 is sent to the face area extraction unit 11.

The face area extraction unit 11 performs template matching between the input image 20 captured from a video camera or the like by the image input unit 10 and a standard face image prepared in advance, as shown in FIG. The face area 21 is extracted from the input image 20. The extracted signal of the face area 21 is sent to the eye / mouth detection unit 12.

The eye / mouth detecting section 12 performs contour extraction processing, binarization processing, and the like on the face area 21 extracted by the face area extracting section 11, and as shown in FIG. Eyes,
Find the center position of the mouth image. In the example of FIG. 3, the center positions 22, 23 of the left and right eyes and the center position 24 of the mouth are indicated by crosses, respectively. The information on the center position of the eye and the mouth detected by the eye / mouth detection unit 12 and the signal of the face area 21 are sent to the affine transformation unit 13.

The affine transformation unit 13 includes an eye / mouth detection unit 12
The affine transformation is performed on the image of the face area 21 so that the center positions of the eyes and the mouth obtained in step 2 are at predetermined positions and the size of the face image is the predetermined size. That is, the affine transformation unit 13 generates a face image 25 whose size and position are normalized from the image of the face area 21 as shown in FIG. The signal of the face image 25 whose size and position have been normalized by the affine transformation unit 13 is sent to the shading removal unit 14.

The shading removal unit 14 removes shading from the face image 25 whose size and position have been normalized by the affine transformation unit 13 by obtaining the shading amount of the image brightness as follows. Do.

Here, the origin is located at the upper left of the calculation area (the face image shown in FIG. 4), the horizontal direction is the x axis, the vertical direction is the y axis, the luminance value of each pixel is the z axis, and the coordinates (x _n , y _n). the luminance values of pixels) and z _n.

First, parameters a, b, and c at which the value of equation (1) is minimum in the calculation area are obtained. At this time, the equation (1) is partially differentiated to obtain “∂f / ∂a = 0” and “∂f / ∂b”.
= 0 ”and“ ∂f / ∂c = 0 ”, the parameters a, b, and c are determined by solving the simultaneous equations.

F (a, b, c) = Σ (z _n- (a * x _n + b * y _n + c)) ² (1) Next, the value of each pixel is calculated by the equation (2). _{Let the} value be the new luminance value z'n.

Z ' _n = z _n- (a * x _n + b * y _n + c) (2) Thus, the shading of the face image 25 is removed. The shading amount obtained by the shading removing unit 14 and the signal of the face image 25 are sent to the contrast correcting unit 15.

The contrast corrector 15 corrects the contrast of the face image 25 by increasing or decreasing the luminance value of the image based on the shading amount obtained by the shading remover 14.

That is, in the contrast correction section 15,
(A * x _n + b * y) obtained by the shading removal unit 14
the average value ave _s of luminance in the face image 25 of the _n + c), further the coordinates _(x n, new luminance value of the pixel of _{y n)
Let} z ″ _{n be} the value of equations (3) and (4). Where d
Is an appropriate constant (where d> 0).

Z '' _n = z ' _n * d * | (ave _s- (a * x _n + b * y _n + c)) | (ave _s- (a * x _n + b * y _n + c )) <0 (3) z '' _n = z ' _n / (d * | (ave _s- (a * x _n + b * y _n + c)) |) (ave _s- (a * x when _{n + b * y n + c} )) ≧ 0 (4) Thus, the contrast correction of the face image 25 is performed. The signal of the face image 25 whose contrast has been corrected by the contrast correction unit 15 is sent to the dispersion processing unit 16.

The distribution processing section 16 and the subsequent average value correction section 17 convert the luminance value so that the average value and the variance of the luminance become predetermined values.

That is, the dispersion processing unit 16 obtains the average value ave ₀ of the luminance with respect to the face image 25 whose contrast has been corrected by the contrast correction unit 15 according to the equation (5), and calculates the variance val _{0 according} to the equation (6). Is calculated, and the value of the equation (7) is calculated for the luminance value z ″ _{n of} each pixel so that the variance becomes a preset variance value val ₁ to obtain a new luminance value z ′ ″ _n .

Ave ₀ = Σz _{2n (5) val 0 = Σ (} z 2n -ave 0) 2 (6) z '''n = (z''n -ave 0) * (val 1 / val 0) 1/2 + ave 0 (7) The variance value obtained by the variance processing unit 16 is sent to the average value correction unit 17.

The average value correction unit 17 applies a luminance value of each pixel to the face image 25 whose variance value has been converted to val ₁ by the variance processing unit 16 so that a predetermined average value ave ₁ is obtained. Ave ₁ −ave ₀ to obtain a new luminance value z ″ ″ _n .

As described above, the size and the position of the image are normalized by the image normalizing apparatus according to the present embodiment.
The output of a face image signal in which the contrast of the dark part of the image is enhanced, while the contrast of the bright part of the image is suppressed, and the change in contrast due to the light and dark of the image is suppressed (contrast is normalized) become.
That is, the face image output from the image normalization apparatus of the present embodiment may be used to perform authentication using the face image, to cause the computer to identify an individual from the face image, or to recognize the expression or gesture of a person. In this case, the face image under a certain condition, in which changes in size, position, illumination, and the like, which are necessary in such a case, are absorbed.

Finally, the description of the above embodiment is an example of the present invention. For this reason, the present invention is not limited to the above embodiments, and various changes can be made according to the design and the like as long as the technical idea according to the present invention is not deviated. It is.

For example, the image area in which the change in contrast can be suppressed by extracting from the input image is not limited to the area of the human face image described in the present embodiment, but may be the area of the animal face image, for example. May be various image areas. Further, the image may be an image of a part of the face such as only the eyes or only the mouth.

[0033]

According to the first aspect of the present invention, there is provided an image normalizing apparatus, comprising: means for obtaining a shading amount of luminance of an image; means for increasing / decreasing a luminance value based on the shading amount; And means for converting the luminance value so that the variance becomes a preset value, thereby enhancing the contrast of the dark part of the image, while suppressing the contrast of the bright part of the image, Contrast change can be suppressed (contrast normalization), and as a result, for example, when authentication is performed using a face image, a computer is used to identify an individual from a face image, or a person's facial expression or gesture is recognized. In such a case, it is possible to obtain a face image under a certain condition that is necessary.

According to a second aspect of the present invention, in the image normalizing method, a shading amount of luminance of an image is obtained, a luminance value is increased or decreased based on the shading amount, and an average value and a variance of luminance are set in advance. By converting the luminance value so that it becomes the same value, the contrast of the dark part of the image is enhanced,
On the other hand, it is possible to suppress the contrast of a bright part of an image and suppress a change in contrast due to the brightness of the image (contrast normalization). As a result, for example, when authentication is performed using a face image, or when a computer is used. It is possible to obtain a face image under a certain condition, which is necessary for identifying an individual from a face image or recognizing a person's expression or gesture.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of an image normalization device according to an embodiment of the present invention.

FIG. 2 is a diagram used to describe an input image and a face area extracted from the input image.

FIG. 3 is a diagram used to explain a center position of an image of an eye and a mouth detected from a face area.

FIG. 4 is a diagram used to describe a face image normalized by affine transformation.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 ... Image input part, 11 ... Face area extraction part, 12 ... Eye and mouth detection part, 13 ... Affine conversion part, 14 ... Shading removal part, 15 ... Contrast correction part, 16 ... Dispersion correction part, 17 ... Average value correction Part, 20: input image, 21: face area, 22, 23: center position of eyes, 24: center position of mouth,
25 ... Face image

Claims (2)

[Claims] 1. A shading calculating means for calculating a shading amount of luminance of an image, an increasing / decreasing means for increasing / decreasing a luminance value based on the shading amount, and an average value and a variance of luminance are set to predetermined values.
A conversion unit for converting the luminance value from the increasing / decreasing unit. 2. A first step for obtaining a shading amount of luminance of an image, a second step of increasing / decreasing a luminance value based on the shading amount, and an average value and a variance of luminance become predetermined values. like,
And a third step of converting the luminance value formed in the second step.