Face recognition through a novel indexing method based on permutations

In face recognition systems that work with databases of thousands of images it is not practical to compare the query image against each stored image to determine similarity. The increasing use of these systems in various fields creates the need to explore mechanisms for efficient and effective search in terms of use of computing resources and percentage of hits. In order to reduce the load of face recognition systems and improve their response time, several alternatives to reduce or eliminate the need for an exhaustive search of images were developed. Indexing methods are one of these alternatives. This paper presents a new indexing technique based on permutations which, optimizes storage of images and accelerates the search process to predict the similarity between objects. The proposed method shows improved behaviour when compared against other techniques representing the state of the art using the FERET database.