Abstract: Recognition in uncontrolled situations is one of the most bottlenecks for practical face recognition. It is crucial to seek a valid and power discriminant method for facial appearance. Based on the local binary patterns for extract local texture, introducing a improved-new method called local ternary patterns, which is more rubost and more power discriminant for illumination variations and noise. Finally, PCA and Fisher linear discriminant arc used to reduce the dimensionalily and optimize discriminative classification respectively. Combing a simple and efficient image preprocessing chain,The mothod is tested and evaluated not only on JDL datasets but also AR datasets, promising results presented the method is valid and feasible.

Key words: Texture extract,Fisher linear discriminant,Local ternary patterns,Filter