Face recognition is an important and fundamental problem in computer vision, and there have been many attempts to address it. Face images have a correlation and high information redundancy which result in inefficiencies instead of such images are used directly for recognition. In this paper, an efficient new hybrid approach to face recognition is presented, which combines image compression and neural network (NN) techniques together. The compression is achieved by applying a way of information packing and fast algorithms to compute discrete cosine transforms (DCTs), to reduce image information redundancy and extracting only the most relevant information that are necessary to preserve the most important facial features. The compressed transform coefficients are used for the system arbitrates between multiple neural networks to improve performance over a single network. The outputs from multiple recognition must be combined into a single decision about the face for classification. The proposed method of face recognition based on DCT much faster than other approaches, by using a very small proportion of transform coefficients. The proposed system is implemented and tested using gray-scale images contains 40 distinct persons, each person having 10 different images and it gives a high recognition rate at high speed.
أ د خميس عواد زيدان | Khamis A. Zidan | 10442