Study of the Pre-processing Impact in a Facial Recognition System

Abstract

The present work is a study of the influence of the preprocessing stage on the classification performance of a face recognition analysis. To carry out this task have made tests in a full FRS, evaluating each of its four stages and including several advanced alternatives in preprocessing, such as illumination normalization through the Discrete Cosine Transformation or alignment by Enhanced Correlation Coefficient, among others. The main goal of this work is determining how those different preprocessing alternatives interact with each other and in wich degree they affect the overall Facial Recognition Systems (FRS). The tests make a special emphasis in using images that could have been obtained from a real environment, rather than at a lab environment, with the difficulties that this brings for facial recognicion techniques.