Procedure of feature extraction from face image for biometrical system

Piotr Kowalski


The aim of this paper is to present a complete procedure for extracting features from the real face picture. The procedure consists of several image processing algorithms such as binarization, erosion, filtration, gradient method and Haar-like Features algorithm. The obtained feature vector used in biometric identification with particular emphasis on the problem of authentication. A system processing the image obtained from a webcam operates in an on-line regime. The correctness of the proposed algorithm was tested using the database created with dozens of people. The tests were done from the camera images as well as artificially created images with different damage cases. The algorithm was tested for cases of various types of lighting, disturbances associated with the natural change of face appearance, the image rotation and distorted image.

Słowa kluczowe: biometrics, authentication, processing and image analysis, image filtering, gradient methods, Haar-like features, classification, neural network