JPH04101280A - Face picture collating device - Google Patents

Abstract

PURPOSE:To obtain a technique which can stably identify individual persons by using their face pictures by constituting the title device of a collating area segmenting means which segments areas required for collation by performing position converting operation on an input picture on the basis of the calculated results of a collating area position calculating means. CONSTITUTION:A collating area position segmenting section 55 reads out a face picture stored in a pattern memory 51 and segments collating areas in accordance with the positional information of the collating areas calculated by means of a collating area position calculating section 54. A feature extracting means 6 prepares a collating pattern by performing a feature extracting process for extracting features to be compared for collation on the normalized and segmented face pictures for collation sent from a position normalization processing means 5. Then a collation processing means 7 collates the collating pattern with standard patterns in a previously prepared identification dictionary file 8 and digitizes the degree of similarity between both patterns. Therefore, individual persons can be identified easily by using their face pictures, since the position and size of front face pictures can be normalized stably.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a face image matching device that performs personal identification using a face image, and in particular, (1) matching a face image of a person registered in advance with a standard pattern. Identity verification to determine whether the person is the same person, (2)? ! [This invention relates to a person search device that selects as a candidate a person who is judged to be most similar to an input image from among a number of registered standard patterns of face images of people.

[Prior art]

Conventionally, human face recognition and classification using two-dimensional images has been performed using pattern matching methods after normalizing facial alignment using feature points extracted from the target facial image. ing. At this time, features necessary for face alignment and normalization are extracted by analyzing edge information obtained by binarizing and applying various operators.

The technology related to the face image matching device is described on pages 17 to 24 of "IEICE Technical Research Report" published by the Institute of Electronics, Information and Communication Engineers, July 13, 1990.

[Problem to be solved by the invention]

However, in the conventional method, feature extraction cannot be performed stably, face alignment and normalization cannot be performed accurately, and as a result, recognition and classification cannot be performed successfully.

An object of the present invention is to provide a technique that allows stable personal identification using facial images.

The above and other objects and novel features of the present invention will become apparent from the description of this specification and the accompanying drawings.

[Means to solve the problem]

In order to achieve the above object, the present invention provides an image input means for inputting a face image of a target person, a density conversion means for performing gradation processing such as brightness correction of the input face image, and an image input means for performing shading processing such as brightness correction of the input face image. A position normalization processing means for normalizing the position and size, a feature extraction means for extracting a feature pattern necessary for matching from the normalized face image, and a feature extraction means for extracting the extracted feature pattern from the normalized face image. It is equipped with a matching processing means that performs matching processing with a set standard pattern, a judgment processing means that judges whether the matching result is valid, and a control means that communicates and controls each processing unit, and matches a person's face image. A face image matching device that realizes personal identification by performing personal identification, wherein the position normalization processing means includes a region division/labeling processing means for automatically extracting facial features by dividing an input image into regions and labeling the input image, and a face image matching device that performs personal identification by A reference point extraction means for extracting a reference point for normalization from the region, a matching region position calculation means for calculating the position and size of a matching region based on the obtained reference points, and a matching region position calculation means for extracting a reference point for normalization from the area of The most important feature is that it is comprised of a matching area cutting means that performs a position conversion operation and cuts out an area necessary for matching.

[Effect]

According to the above-mentioned means, when performing personal identification using human face images, edge information, which has not been used in the past, is used to extract facial reference points necessary for normalizing position and size. Rather than extracting feature points from the image, we can stably normalize the position and size by finding the areas corresponding to facial features such as eyes and mouth, which can be extracted more stably. Therefore, it is possible to stably identify individuals based on facial images.

[Embodiments of the invention]

Hereinafter, one embodiment of the present invention will be specifically described using the drawings.

In addition, in all the figures for explaining the embodiment, parts having the same functions are given the same reference numerals, and repeated explanations thereof will be omitted.

In inputting a frontal image of a face, the present invention stably determines the area to be subjected to subsequent feature extraction and comparison processing with an identification dictionary, regardless of changes in the distance from the camera to the subject's face or movement of the face. For extraction, the matching area can be freely set using parameters based on the reference point. For example, the following matching area can be considered.

(1) Area around both eyes (2) Area on the face including both eyes, nose, and mouth (3) Area on the entire face including hair In this example, as an example, the area around both eyes in (1) The case of matching will be explained.

FIG. 1 is a block diagram showing a schematic configuration of an embodiment of the face image matching device of the present invention, and FIG. 2 is a block diagram showing the configuration of a functional system of the position normalization processing means in FIG. 1. be.

In FIG. 1, 1 is a person to be recognized or classified, 2 is a television camera, 3 is an image input means, 4 is a density conversion means, 5 is a position normalization processing means, 6 is a feature extraction means, and 7 is a Reference numeral 8 denotes an identification dictionary file, 9 denotes a determination processing means, and 10 denotes a control means for managing the progress of the entire process.

Note that here, the image input means 3. Concentration conversion means 4,...
- The control means 10 is a component that can be constructed within the same computer.

The position normalization processing means 5, as shown in FIG. ing.

Next, the operation of the face image matching device of this embodiment will be explained.

The face of a target person 1 is photographed by a television camera 2 shown in FIG. At this time, the photographed image becomes a shading or color image through subsequent processing. The photographed face image is sent to the image input means 3.

The image input means 3 converts the image sent from the television camera 2 into a format suitable for subsequent processing, and sends it to the density conversion means 4. The density conversion means 4 performs density conversion processing such as brightness value correction (illuminance value normalization) on the face image input by the image input means 3, and sends it to the position normalization processing means 5. The position normalization processing means 5 normalizes the position and size of the sent face image, and then performs a process of cutting out a region necessary for verification.

The processing here will be explained along with FIG.

The face image sent to the image input means 3 is input to the pattern memory 51, the face image is stored until the next input, and the stored image is provided as needed.

Next, the area division/labeling processing unit 52 reads out the face image stored in the pattern memory 51.
This read facial image is subjected to region segmentation and labeling to extract each feature region and information of the face.

As for the area segmentation and labeling used here, for example, it has been confirmed that the area segmentation/labeling method that focuses on image color information by Wallace et al. works effectively for facial images (Wallace, Suenaga et al. rMDL
“Region Segmentation of Color Facial Images Using Clustering”, IEICE University, 5D-11-3, 1990-6, and Ri
chard S, Wallace Takeo, Kan
ade “Finding natural cl.
usters having minimum
description length,”10th
ICPR1990, 1990-6).

Region segmentation of facial images by Wallace et al.
The results of the labeling process are shown in Figure 3.

Next, the reference point extraction unit 53 aligns the facial image based on the position information of each feature of the face obtained by the region segmentation/labeling processing unit 52, and determines reference points for normalizing the size. Extract. Specifically, based on the results of the reference point extraction unit 53, for example, both eyes.

Er, El, and M can be extracted by using the center of gravity of the mouth region or a histogram along the vertical or horizontal axis. For convenience, these three points (Er.

El, M) as the reference point.

The matching area position calculating unit 54 calculates the position of the area required for matching based on the reference points obtained by the reference point extracting unit 53. Specifically, a method as shown in FIG. 4 can be considered, which is carried out in the following steps.

(1) Leg O of the perpendicular line drawn from M to the straight line connecting Er and El
shall be.

(2) Parallel and rotational movement is performed so that the line segment OM coincides with the perpendicular line of the image, and its -indoor segment point is at a fixed position.

(3) The image is isotropically enlarged or reduced so that the length of OM-g becomes a constant value.

The matching area cutting unit 55 reads and calls the face image stored in the pattern memory 51) and cuts out the matching area according to the positional information of the matching area calculated by the matching area position calculating unit 54. The cut out matching area is sent to the feature extraction means 6.

The feature extracting means 6 performs extraction processing on the normalized and extracted face image for verification sent from the position normalization processing means 5 to extract features to be compared during verification, thereby generating a verification pattern.

Next, in the matching processing means 7, the feature pattern consisting of the features extracted by the feature extraction means 6 is subjected to the processing up to the feature extraction means 6 and registered in advance in the prepared identification dictionary file 8. , and quantify the similarity measure between the two.

The determination processing means 9 performs threshold processing using the optimum value for the form in which the data group of the similarity measure between the input pattern calculated by the matching processing means 7 and the standard pattern of each category is to be used. It makes a judgment and outputs the result.

As can be seen from the above description, according to this embodiment, when performing personal identification using a person's face image, the conventional Rather than extracting feature points based on edge information, which had previously been used, the normalization of position and size is achieved by extracting feature points from areas corresponding to facial features such as eyes and mouth, which can be extracted more stably. This allows stable identification of individuals based on facial images.

In the above embodiment, the present invention has been described with reference to the case where the area around both eyes is used as the matching area, but it goes without saying that the invention can be applied to other matching areas.

The present invention has been specifically explained above based on examples, but
It goes without saying that the present invention is not limited to the embodiments described above, and can be modified in various ways without departing from the spirit thereof.

〔Effect of the invention〕

As described above, according to the present invention, it is possible to stably normalize the position and size of a face image input from the front, and therefore it is possible to stably perform personal identification using a face image. .

[Brief explanation of drawings]

1 is a block diagram showing a schematic configuration of an embodiment of the face image matching device of the present invention; FIG. 2 is a block diagram showing the configuration of a functional system of the position normalization processing means in FIG. 1; The figure shows the results of region segmentation and labeling processing of a face image by Wallace et al.
FIG. 3 is a diagram illustrating an example of a method for normalizing a face image. In the figure, 1...Target person, 2...TV camera, 3.
・Image input means, 4: Density conversion means, 5: Position normalization processing means, 6: Feature extraction means, 7: Verification processing means, 8: Identification dictionary file, 9: Judgment Processing means, 10... Control means, 51... Pattern memory, 5
2... Area division/labeling processing unit, 53... Reference point extraction unit, 54... Verification area position calculation unit, 55...
Verification area extraction part. 10th

Claims (1)

[Claims] (1) Image input means that performs input processing of a face image of a target person, density conversion means that performs shading processing such as brightness correction of the input face image, and normalization of the position and size of the input face image. A position normalization processing means, a feature extraction means that performs a process of extracting a feature pattern necessary for matching from a normalized face image, and a process of matching the extracted feature pattern with a standard pattern prepared in advance. Verification processing means to perform;
A face image matching device that realizes personal identification by matching face images of a person, comprising a determination processing means for determining whether a matching result is valid, and a control means for communicating and controlling each processing unit. , the position normalization processing means includes region division/labeling processing means for automatically extracting facial features by dividing and labeling the input image into regions, and reference point extraction means for extracting reference points for normalization from the extracted feature regions. , a matching area position calculation means that calculates the position and size of the matching area based on the obtained reference point, and a matching unit that performs a position conversion operation on the input image based on the calculation result and cuts out the area required for matching. A face image matching device comprising a region cutting means.