Using principles of fractal image compression for complexity estimation of the face recognition problem

Abstract

An approach to solve the problem of identifying individuals in static or dynamic images has been presented. The given approach is based on the methods of a fractal image compression. Algorithms for template image and current image processing have been proposed. For the proposed algorithms the computational complexity is estimated. In contrast with many other recognition algorithms the complexity of the proposed image processing scheme has linear dependence on the size of the image. The disadvantages of this algorithm are high complexity of the template image processing and linear dependence of the recognition speed on the size of the face database. Nevertheless, this algorithm can be effectively parallelized, e.g. with computing on GPU.