KR19990079823A - Face Recognition Method Considering Lighting Change - Google Patents

Face Recognition Method Considering Lighting Change Download PDF

Info

Publication number
KR19990079823A
KR19990079823A KR1019980012628A KR19980012628A KR19990079823A KR 19990079823 A KR19990079823 A KR 19990079823A KR 1019980012628 A KR1019980012628 A KR 1019980012628A KR 19980012628 A KR19980012628 A KR 19980012628A KR 19990079823 A KR19990079823 A KR 19990079823A
Authority
KR
South Korea
Prior art keywords
face
ρ
image
gallery
normal
Prior art date
Application number
KR1019980012628A
Other languages
Korean (ko)
Other versions
KR100287216B1 (en
Inventor
기석철
Original Assignee
윤종용
삼성전자 주식회사
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 윤종용, 삼성전자 주식회사 filed Critical 윤종용
Priority to KR1019980012628A priority Critical patent/KR100287216B1/en
Publication of KR19990079823A publication Critical patent/KR19990079823A/en
Application granted granted Critical
Publication of KR100287216B1 publication Critical patent/KR100287216B1/en

Links

Abstract

The present invention relates to a face recognition method in which illumination changes are considered, and a face recognition method that considers illumination changes is a method of recognizing a face by receiving an image of faces (galleries) ρ A normal registration step of calculating and storing an average value and a surface normal n s of the image; Obtaining a normal average of all faces of the gallery; Face image brightness, ρ Estimating a direction of light exposed at the time of capturing an input face image by using an average value and a normal mean of the input image; Estimated light direction and registered gallery ρ And calculating a brightness of each gallery using a normal line; And comparing the calculated brightness of each gallery and the brightness of the face of the newly input picture to find the most similar gallery to recognize the face.
According to the present invention, by considering the change of illumination, the accuracy of face recognition is increased at a place where illumination change is severe, and the normal of the face can be extracted to easily apply to the face synthesis field.

Description

Face Recognition Method Considering Lighting Change

[0001] The present invention relates to a face recognition system, and more particularly, to a face recognition system which extracts and registers parameters considering illumination change from each face image constituting a gallery to be registered, recognizes an object to be recognized in consideration of illumination, And a face recognition method.

Generally, face recognition technology is applied to verification of identity such as speech recognition, fingerprint recognition, and eye recognition. The face recognition technology can be implemented at a relatively low price compared to other identification verification techniques, and it is possible to prevent the rejection feeling of the recognition object because the user does not need to take special action for recognition. Particularly since the multimedia capture (PC) has been popularized and the image acquisition device is basically installed, it is possible to apply the face recognition technology without a separate hardware configuration.

Conventionally, a technique of recognizing human faces (Yael Adini, Yael Moses and Shimon Ullman, "Face Recognition: The Problem of Compensating for Changes in Illumination Direction," IEEE Trans, Pattern Analysis and Machine Intelligence, Vol. , pp721 ~ 732 July 1997.), the recognition parameters and the distance scale, which are least sensitive to the change of illumination, are selectively used when the illumination change is severe. For example, recognition parameters such as an edge map, image intensity derivatives, a 2D Gabor-like filter, And recognizing it as an affine-GL distance measure.

The above equation (1) is a formula for comparing the face image of I 1 and the face image of I 2 to calculate the degree of similarity. Since the face photographs are expressed by the degree of brightness, even if the faces are the same, if the direction of the light or the light at the time of photographing the face is different, it is inevitably recognized as different faces by the above method. That is, the most fundamental problem of face recognition due to illumination change is that only two-dimensional brightness information distorted by illumination can be used in recognizing a face having a three-dimensional shape. There is a problem in that, if the illumination change is severe, the face recognition method that does not take into consideration the conventional illumination change described above can hardly succeed in recognition.

The present invention provides a method of registering a gallery having an average reflection coefficient and a normal vector of a face surface and estimating illumination direction of a face to be recognized using the method, thereby performing face recognition under the same illumination condition as the gallery And to provide a face recognition method considering illumination change.

FIG. 1 is a view illustrating a photographing scheme for obtaining a basis image to be inputted at the time of registering a gallery of the present invention.

FIG. 2 is a flowchart illustrating a method of registering a gallery for face recognition according to the present invention.

3 is a flowchart of a face recognition method according to the present invention.

A method of registering a face image (gallery) as a comparison object to be recognized in face recognition in order to solve the above problem is a method in which a face of a person to be registered is exposed to illumination in three directions, ); The surface reflection coefficient of each pixel forming the face in the basis image ρ ; In each pixel, ρ Calculating an average value for the average value; Calculating a surface normal of each pixel constituting the face; And the calculated ρ And registering and storing the normal line.

remind ρ Is calculated by using the face brightness vector I and the three illumination direction vectors N for three illumination directions, assuming that the face image or image is a Lambertian surface model,

ρ = | N -1 I |

.

The normal calculation of the facial surface can be expressed by the following equation (3)

.

According to another aspect of the present invention, there is provided a face recognition method in which illumination changes are considered. The face recognition method includes receiving an image of faces (galleries) to be registered, ρ A normal registration step of calculating and storing an average value and a surface normal n s of the image; Obtaining a normal average of all faces of the gallery; The face photograph brightness to be recognized, ρ Estimating a direction of light that is exposed at the time of capturing the input face image by using an average value of the face images and the normal line average; The direction of the estimated light and the direction of the registered gallery ρ And calculating a brightness of each gallery using a normal line; And comparing the calculated brightness of each gallery and the brightness of the face of the newly input picture to find the most similar gallery to recognize the face.

Wherein the gallery registration step comprises: receiving a basis image of a face of a person to be registered exposed to illumination in three directions; The surface reflection coefficient of each pixel forming the face in the basis image ρ ; In each pixel, ρ Calculating a surface normal of each pixel forming the face; And the calculated ρ And registering the normal line and storing the normal line.

The step of estimating the direction of the exposed light when capturing the image of the object to be recognized comprises the steps of obtaining the brightness information I (x, y) for each pixel of the photographic image since the object to be recognized is input as a single photograph, ρ Using the mean and the average normal, n s , using the least square error method,

I (x, y) =? Nn s

It is preferable to obtain the direction vector N in a specific illumination.

Wherein the step of calculating the brightness of each of the faces belonging to the gallery further comprises: ρ To be a step of averaging the direction of the illumination vector N and using the normal n s of each of the registered output brightness I, such as the equation (4) is preferred.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will now be described in detail with reference to the accompanying drawings.

FIG. 1 illustrates a basis image photographing scheme for registering a gallery according to the present invention. In order to input a face image of a person, a face image exposed in three different directions, that is, a basis image FIG. Usually, an image called a face is picked up by a CCD camera or the like and processed by a predetermined application program in a computer or a predetermined face recognition dedicated hardware. At this time, since the image of the face image taken is the brightness of the image constituting the image, an error may occur that the images of the same face are recognized as different faces depending on the position of light or illumination at the time of imaging. Due to the possibility of this error, the data of the various facial images to be registered for recognition requires three images each captured under illumination in different positions. From each of these basis images, the mean value and the normal vector of the surface reflection coefficient of the facial surface are extracted and stored and used as a parameter for recognition. That is, an average of normal vectors of a plurality of registered face images (hereinafter, referred to as 'gallery') registered for recognition is searched and used to extract the illumination direction of a face image to be recognized. It is assumed that the normal vector is a vector related to the outline of the face which is not related to the brightness of the face, and that there is not a large difference in the outline of all the faces. As shown in Fig. 1, three positions that are not located on the same plane in a three-dimensional space are preset, and three basis images are acquired for each recognition target face while illuminating the illumination at each position. I 3 (x, y) and I 3 (x, y) are defined as I 1 (x, y), I 2 In this case, the position vectors of each illumination are denoted by n 1 , n 2 , and n 3 , respectively, and defined as a direction matrix N. These vectors are shown in the following equation.

I = [I 1 , I 2 , I 3 ]

n 1 = [n 11 n 12 n 13 ] '

n 2 = [n 21 n 22 n 23 ]

n 3 = [n 31 n 32 n 33 ]

Assuming that the facial image is a lambertian surface model, the brightness I (x, y) in each pixel of the facial image can be calculated by Equation (4). here ρ Is the facial surface reflectance and n s is the surface normal of the face. In the face image, the face surface brightness coefficients are calculated differently according to the positions of the pixels. However, since the outline of the face is a more important recognition parameter for the face recognition, it is assumed that the reflection coefficients of the face surface are all the same for the simplification of calculation. Using the brightness I of the pixel forming the facial image, ρ As shown in Equation (2). This calculated ρ The normal line n s , which is an important parameter for face recognition, is calculated as shown in equation (3).

FIG. 2 is a flowchart of a method for registering a face image for face recognition according to the present invention. In the method for registering a gallery as a database for face recognition, first, face images of a person to be registered are three Three basis images exposed in different illumination directions are input (step 200). The surface reflection coefficient of each pixel that forms the face of the basis image ρ (Step 210). ρ Is calculated by the above-described equation (2). In Equation (2), the direction matrix N of the illumination and its inverse matrix N- 1 can be known because the position of the illumination is fixed and the brightness I of the pixel can be known. ρ Can be easily calculated. In order to simplify the calculation, it is assumed that the reflection coefficients of the facial surface are all the same. The reflection coefficient of all the pixels forming the face under this assumption ρ And the average reflection coefficient value is divided by the total number of pixels (Step 220). A normal line n s at each pixel is calculated using Equation (3) (Step 230). At this time ρ The average calculated in step 220 ρ Value. This calculated ρ The average and normal n s are registered and stored (step 240). Counts the number of faces to be registered, and applies the steps from 200 to 240 for each face to complete the gallery registration. If you want to register 10 faces as a gallery, you need 30 images because there are 3 basis images to be input in 200 steps for one face.

3 is a flowchart of a face recognition method according to the present invention. In the face recognition method for processing a face input to be newly recognized by a face recognition system having a registered gallery as shown in FIG. 2, n s (step 300). When a subject to be recognized (hereinafter referred to as a probe) is input, that is, when a face image to be recognized is input, the brightness I of the face and the average ρ And the average n s , the illumination direction in which the photograph input by the least squares error method is imaged is estimated (step 310). From the above equation (4), the following equation (5) can be derived.

Here, I i is the image, that is, the brightness value of the i-th pixel of the probe to be recognized, m is a gallery in the direction of the light to be estimated is the pixel number, n x of the entire image probe, and n s i 2 FIG. Means the normal at the i-th pixel of the average image calculated at the time of registration. N x can be estimated by applying a least square method to Equation (6). The direction of the estimated light and the direction of the light ρ Using the average and the normal, the brightness of each of the registered galleries is again computed by applying Equation (4) (Step 320). This is to match the illumination conditions of the currently registered probe with the registered gallery. This is because, if the illumination condition of the probe to be recognized and the gallery to be compared are not the same, even the same face may be recognized as different faces because the brightness is different. The brightness of each gallery is compared with the brightness of the probe to be recognized, and the recognition result with the smallest difference is calculated (operation 330). The above-described face recognition method considering illumination change can be applied to a screen saver of a computer or a security system. For example, in the case of a screen saver, a password is inputted or a user input is detected in order to release a screen saver in the existing one. However, after the face of a user of the computer is registered as a gallery using the face recognition method of the present invention, The user can recognize the same face as the registered face by inputting the face of the person who intends to use the computer by using the camera of the computer, and can release the screen saver if the face is the same face. In the case of the security system, the camera is installed in a place requiring security such as automatic car security, office door locking device, etc., and registration of the gallery according to the present invention and face recognition are performed, .

According to the present invention, by considering the change of illumination, the accuracy of face recognition is increased at a place where illumination change is severe, and the normal of the face can be extracted to easily apply to the face synthesis field.

Claims (7)

  1. A method of registering a face image (gallery) serving as a comparison object of an object to be recognized upon face recognition,
    Receiving a captured basis image by exposing a face of a person to be registered to illumination in three directions;
    The surface reflector coefficients of each of the pixels forming the face in the basis image ρ ;
    In each pixel, ρ Calculating an average value for the average value;
    Calculating a surface normal of each pixel constituting the face; And
    Calculated ρ And a step of registering and storing the normal line and the normal line.
  2. The method according to claim 1, ρ Is calculated,
    Assuming that the face photograph or image is a Lambertian surface model, the face brightness vector I for the three illumination directions, and the three illumination direction vectors N,
    ρ = | N -1 I |
    The method of claim 1,
  3. 3. The method according to claim 2,
    As shown in the following equation,
    And calculating a face image based on the face recognition result.
  4. In the face recognition method,
    The image of the face (gallery) to be registered is input, and the surface reflection coefficient ρ A normal registration step of calculating and storing an average value and a surface normal n s of the image;
    Obtaining a normal average of all faces of the gallery;
    The face photograph brightness to be recognized, ρ Estimating a direction of light that is exposed at the time of capturing the input face image by using an average value of the face images and the normal line average;
    The direction of the estimated light and the direction of the registered gallery ρ And calculating a brightness of each gallery using a normal line; And
    And recognizing the face by comparing the brightness of each gallery and the face brightness of the newly input picture to find the most similar gallery.
  5. [5] The method of claim 4,
    Receiving a captured basis image by exposing a face of a person to be registered to illumination in three directions;
    The surface reflection coefficient of each pixel forming the face in the basis image ρ ;
    In each pixel, ρ Calculating a surface normal of each pixel forming the face; And
    Calculated ρ And registering the normals and registering the normals, and storing the registered normals.
  6. 5. The method according to claim 4, wherein the step of estimating the direction of the exposed light at the time of picking-
    Since the recognition object is input as a single photograph, brightness information I (x, y) for each pixel of the photographic image is obtained, ρ Applying the mean and the average normal n s to the expression of the least squares method,
    I (x, y) =? Nn s
    And obtaining a direction vector N of the illumination.
  7. 5. The method of claim 4, wherein calculating the brightness of each of the faces belonging to the gallery further comprises:
    remind ρ Mean the direction of the illumination vector N and a face recognition method considering an illumination change, it characterized in that the step of calculating the brightness I, such as the equation (4) using the normal n s of each of the registration.
KR1019980012628A 1998-04-09 1998-04-09 Robust face recognition system under varying illumination KR100287216B1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1019980012628A KR100287216B1 (en) 1998-04-09 1998-04-09 Robust face recognition system under varying illumination

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1019980012628A KR100287216B1 (en) 1998-04-09 1998-04-09 Robust face recognition system under varying illumination

Publications (2)

Publication Number Publication Date
KR19990079823A true KR19990079823A (en) 1999-11-05
KR100287216B1 KR100287216B1 (en) 2001-01-20

Family

ID=37514972

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1019980012628A KR100287216B1 (en) 1998-04-09 1998-04-09 Robust face recognition system under varying illumination

Country Status (1)

Country Link
KR (1) KR100287216B1 (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100647298B1 (en) * 2004-11-29 2006-11-23 삼성전자주식회사 Method and apparatus for processing image, and computer readable media for storing computer program considering light reflection
KR100756047B1 (en) * 2006-01-25 2007-09-07 한국인식산업(주) Apparatus for recognizing a biological face and method therefor
KR100876786B1 (en) * 2007-05-09 2009-01-09 삼성전자주식회사 System and method for verifying user's face using light masks
KR100893086B1 (en) * 2006-03-28 2009-04-14 (주)코아정보시스템 Method for detecting face robust to illumination change
KR100907597B1 (en) * 2007-12-05 2009-07-14 에스케이 텔레콤주식회사 The lighting control based on facial recognition system and method, and this applied devices, and servers

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101844371B1 (en) 2014-03-19 2018-04-02 삼성전자주식회사 Method and apparatus for processing image

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100647298B1 (en) * 2004-11-29 2006-11-23 삼성전자주식회사 Method and apparatus for processing image, and computer readable media for storing computer program considering light reflection
KR100756047B1 (en) * 2006-01-25 2007-09-07 한국인식산업(주) Apparatus for recognizing a biological face and method therefor
KR100893086B1 (en) * 2006-03-28 2009-04-14 (주)코아정보시스템 Method for detecting face robust to illumination change
KR100876786B1 (en) * 2007-05-09 2009-01-09 삼성전자주식회사 System and method for verifying user's face using light masks
KR100907597B1 (en) * 2007-12-05 2009-07-14 에스케이 텔레콤주식회사 The lighting control based on facial recognition system and method, and this applied devices, and servers

Also Published As

Publication number Publication date
KR100287216B1 (en) 2001-01-20

Similar Documents

Publication Publication Date Title
Breitenstein et al. Real-time face pose estimation from single range images
EP1800259B1 (en) Image segmentation method and system
Kollreider et al. Verifying liveness by multiple experts in face biometrics
US6188777B1 (en) Method and apparatus for personnel detection and tracking
US7412081B2 (en) Personal authentication apparatus and personal authentication method
JP3975248B2 (en) Biometric recognition using neural network classification
DE60123378T2 (en) Digital image processing method with different ways of detecting eyes
US9087232B2 (en) 3D object recognition
JP4755202B2 (en) Face feature detection method
US8345921B1 (en) Object detection with false positive filtering
Davis et al. Robust Background-Subtraction for Person Detection in Thermal Imagery.
US7218759B1 (en) Face detection in digital images
US8885074B2 (en) Red-eye filter method and apparatus
JP3954484B2 (en) Image processing apparatus and program
De Carvalho et al. Exposing digital image forgeries by illumination color classification
US8331619B2 (en) Image processing apparatus and image processing method
JP4451307B2 (en) How to classify objects in a scene
CA2571643C (en) Single image based multi-biometric system and method
EP1693783B1 (en) Fast method of object detection by statistical template matching
US8836777B2 (en) Automatic detection of vertical gaze using an embedded imaging device
JP5008269B2 (en) Information processing apparatus and information processing method
Fattal Dehazing using color-lines
US7912252B2 (en) Time-of-flight sensor-assisted iris capture system and method
KR20080095999A (en) Recognition system
US7801335B2 (en) Apparatus and methods for detecting the presence of a human eye

Legal Events

Date Code Title Description
A201 Request for examination
E902 Notification of reason for refusal
E701 Decision to grant or registration of patent right
GRNT Written decision to grant
FPAY Annual fee payment

Payment date: 20090112

Year of fee payment: 9

LAPS Lapse due to unpaid annual fee