Face data base compression by hotelling transform using segmentation

Face images could be projected onto a feature space that the variation among them can be expressed better. The face space is defined by the ldquoeigenfacesrdquo, which are the eigenvectors of the faces set; they do not necessarily correspond to the isolated features such as eyes, ears, and noses. Using eigenfaces decreases the huge amount of the required memory to save face database. In this paper we use the similarity in different segments of faces. So, we divide the face into some special blocks and apply hotelling transform to each block separately. Simulation results show a decrease of 40% in the compression ratio. The mentioned algorithm has also the ability of processing in parallel.