Improved preprocessing techniques for illumination-invariant face recognition

Abstract

Illumination Variation is one of the main obstacles in automatic face recognition systems. In the past few years, many approaches to coping with illumination variations have been proposed which can be categorized into model based and preprocessing based. Although the model-based approaches seem more perfect in theory, they commonly introduce more constraints and computational time is very prohibitive in most cases. On the other hand, the preprocessing approaches commonly exploit simple and efficient image processing techniques. The typical approaches based on image processing include histogram equalization (HE), histogram specification (HS), derivatives of the gray-level images by Laplacian of Gaussian filter (LOG), wavelet based method, and quotient image bases methods like self quotient image (SQI). In this thesis, I propose some improved methods based on the LOG and wavelet based method. The main idea is to apply image de-noising to suppress noises and Gamma Adjustment to alleviate the uneven illumination conditions. After that we extract and enhance facial features using LOG and wavelet decomposition so as to facilitate our recognition task. By comparing proposed methods with previous methods on the CAS-Peal Face database and Extended Yale Face Database B, proposed methods showed an obvious improvement on both databases compared with previous major methods which showed excellent performance. Experimental results proved the effectiveness of the proposed methods. Our proposed methods do not need any training or time consuming optimization procedure, so they are suitable for real time applications.