Face Recognition Across Non-Uniform

Motion Blur, Illumination & Pose

Contents

Biometric, its types Face recognition Advantages Across Non-Uniform Motion Blur, Illumination, and Pose Template creation Nodal points Process Modules Conclusion Reference

What is meant by Biometrics?

A biometric is a unique, measurable characteristic of a human being that can be used to automatically recognize an individual or verify an individual’s identity.

Biometrics can measure both physiological and behavioral characteristics.

PHYSIOLOGICAL

a. Finger-scan b. Facial Recognition c. Iris-scan d. Retina-scan e. Hand-scan

BEHAVIORAL

a. Voice-scan b. Signature-scan c. Keystroke-scan

CONTENT

Why Face Recognition… Everyday actions are increasingly being handled electronically,

instead of pencil and paper or face to face. This growth in electronic transactions results in great demand

for fast and accurate user identification and authentication. Access codes for buildings, banks accounts and computer

systems often use PIN's for identification and security clearances.

Using the proper PIN gains access, but the user of the PIN is not verified. When credit and ATM cards are lost or stolen, an unauthorized user can often come up with the correct personal codes.

Face recognition technology may solve this problem since a face is undeniably connected to its owner expect in the case of identical twins.

Advantages of using Face Recognition It requires no physical interaction on behalf of the

user.

It is accurate and allows for high enrolment and verification rates.

It can use your existing hardware infrastructure, existing cameras and image capture devices, with no problems.

Face Recognition… How?

In Facial recognition there are two types of comparisons:-

VERIFICATION- The system compares the given individual with who they say they are and gives a yes or no decision.

IDENTIFICATION- The system compares the given individual to all the Other individuals in the database and gives a ranked list of matches.

All identification or authentication technologies operate using the following four stages:

Capture: A physical or behavioral sample is captured by the system during Enrollment and also in identification or verification process.

Extraction: unique data is extracted from the sample and a template is created.

Comparison: the template is then compared with a new sample.

Match/non-match: the system decides if the features extracted from the new samples are a match or a non match.

Across Non-Uniform MotionBlur, Illumination, and Pose Enrolment templates are normally created from a

multiplicity of processed facial images. These templates can vary in size from less than 100

bytes, generated through certain vendors and to over 3K for templates.

The 3K template is by far the largest among technologies considered physiological biometrics.

Larger templates are normally associated with behavioural biometrics.

Template creation

How Facial Recognition System Works

Facial recognition software is based on the ability to first recognize faces, which is a technological feat in itself. If you look at the mirror, you can see that your face has certain distinguishable landmarks. These are the peaks and valleys that make up the different facial features.

There are about 80 nodal points on a human face.

Nodal points used

Here are few nodal points that are measured by the software.

1. distance between the eyes 2. width of the nose 3. depth of the eye socket 4. cheekbones 5. jaw line 6. chin

Detection- when the system is attached to a video surveilance system, the recognition software searches the field of view of a video camera for faces. If there is a face in the view, it is detected within a fraction of a second. A multi-scale algorithm is used to search for faces in low resolution. The system switches to a high-resolution search only after a head-like shape is detected.

Alignment- Once a face is detected, the system determines the head's position, size and pose. A face needs to be turned at least 35 degrees toward the camera for the system to register it.

Face recognition systems that work with focused images have difficulty when presented with blurred data. Approaches to face recognition from blurred images can be broadly classified into four categories. (i) Deblurring-based in which the probe image is first deblurred and

then used for recognition. However, deblurring artifacts are a major source of error especially for moderate to heavy blurs.

(ii) Joint deblurring and recognition, the flip-side of which is computational complexity.

(iii) Deriving blur-invariant features for recognition. But these are effective only for mild blurs.

(iv) The direct recognition approach in which reblurred versions from the gallery are compared with the blurred probe image.

It is important to note that all of the above approaches assume a simplistic space-invariant blur

model. For handling illumination, there have mainly been two directions of pursuit based on

(i) the 9D subspace model for face and (ii) extracting and matching illumination insensitive facial features

(a) Focused image,

(b) synthetically blurred image obtained byapplying random in-plane translations and rotations on the focused image,

(c) point spread functions (PSF) at various locations in the image showing thepresence of non-uniform blur which cannot be explained by the convolutionmodel

and (d, e, f) real blurred images

Process Normalization-The image of the head is scaled and rotated so

that it can be registered and mapped into an appropriate size and pose. Normalization is performed regardless of the head's location and distance from the camera. Light does not impact the normalization process.

Representation-The system translates the facial data into a unique code. This coding process allows for easier comparison of the newly acquired facial data to stored facial data.

Matching- The newly acquired facial data is compared to the stored data and (ideally) linked to at least one stored facial representation.

Modules

Motion blur model for faces Face recognition across blur Multiscale implementation Recognition across blur, illumination and pose

Modular description

Motion blur model for faces The apparent motion of scene points in the image will

vary at different locations. The single blur kernel can’t explain this. In the proposed method a space variant motion blur

model is presented and the explanation for geometric degradations of faces resulting from camera motion is illustrated.

An optimization algorithm to recover the camera motion is proposed.

Multiscale implementation The difference in the displacement of a point light source

due to two different transformations from the discrete set T is at least one pixel.

Doubling the sampling resolution increases the total number of poses.

Face recognition across blur We have M face classes with one focussed face fm for

each class m, m=1,2,… ,M. A convex combination of these are then produced.

Recognition across blur, illumination and pose Finally pose variation is allowed, in addition to blur and

illumination. Four near frontal poses are selected, angles within ~15

degree and an algorithm called MOBIL is applied in small variations in pose

A correct pose is then obtained.

Conclusion

A methodology to perform face recognition under the combined effects of non-uniform blur, illumination, and pose is proposed. The set of all images obtained from a given image by non-uniform blurring and changes in illumination forms a bi-convex set, this result is used to develop non-uniform motion blur and illumination-robust algorithm called MOBIL.

The limitation of his approach is that significant occlusions and large changes in facial expressions cannot be handled.

References

• W. Zhao, R. Chellappa, P. J. Phillips, and A. Rosenfeld, “Face recognition: A literature survey,”

• http://www.nec.com/en/global/solutions/biometrics/technologies/face_recognition.html

Thank you