ENHANCED FACE RECOGNITION USING ILLUMINATION VARIATIONS BASED ON LOCAL BINARY PATTERN

Abstract

A new face recognition algorithm proposed by Z.Lian, M.J.Er and J.Li is implemented in this research paper. Unlike the other face recognition schemes, the proposed algorithm works efficiently in varying Illumination conditions. This paper highlights the limitation of present face recognition algorithms. Algorithms offering efficient feature extraction methods fail to differentiate between similar images under varying light conditions. Algorithms with efficient preprocessing methods give a lot of errors during distance computation. Distance measurement such as Histogram Intersection works well at global level but fails to compute pixel distance effectively. Similarly hamming distance gives better approximation of pixel distance but performs poorly at global level. A solution to the highlighted problems is presented in terms of new proposed algorithm by Z.Lian, M.J.Er and J.Li. The new algorithm is explained with the help of block diagram and experimental results. A detailed testing is done on Yale B & Extended Yale B data sets and further comparison with other existing schemes is also included to highlight the advantages of proposed method.