Face Classification by Local Texture Analysis through CBIR and SURF Points

This study presents a robust face recognition system that takes into account both, local texture and points-of-interest analysis. This system uses the CBIR (Content Based Image Retrieval) technique considering as descriptors the mean, the standard deviation, and the homogeneity of each of the several image windows subjected to analysis; that is, each window acts as a local image region subjected to the face analysis having a face point of interest at its center. In this way, the system retrieves descriptive data of people by analyzing their own texture characteristics on the interior of each face. The system achieves to get a self-organization of the data which a similarity-based order approximation of the face images in a database (DB). With the support of the analysis provided by the points of interest technique SURF, complemented with the CBIR technique, we generated a robust map able to achieve a 100% classification conducted on DB. The results have also been highly successful when conducted under controlled lighting conditions.