The Research For Fractal Feature Extraction And Face Recognition Algorithm

Since face is one of the most important features to distinguish among objects, the research of face recognition technology becomes significant to computer vision and pattern recognition. Feature extraction and pattern classification are important steps in the face recognition system, so the stable feature is related to the result of recognition. The system efficiency depends on the choice of classifier largely.In this paper, according to self-similarity characteristics of face image, we proposed a plan based on fractal feature, then propose fast fractal feature extraction algorithm based on the symmetric properties of the face's left and right, finally combine to hidden markov model to achieve face recognition. Specific work is as the following:Firstly, by analysis the self-similarity and the symmetric properties of the face's left and right, we proposed a fast and effective plan based on fractal feature, then use the obtained fractal feature as the basis of the judgment to the face recognition system, finally using the method of fractal distance to distinguish. Experiment results show that the algorithm is with the advantage of quick response on matching and high accuracy in recognition and has the good performance of recognition.Calculate the matrix by using extracts fractal features for the above mentioned. Firstly, we do the Block matrix with the fractal features and by fuzzy clustering analysis to get HMM observation sequence. What's more, the HMM is being trained and established in the process of face recognition. In the end, the trained HMM could be used to identify faces. Experiment results support that this algorithm has high recognition rate and good perspective.