Research On Face Recognition Algorithm Based On Fractal Theory

Fractal transform theory has been obtain a series of successful application in natural science,philosophy and social science.Applying fractal theory to face recognition is a new method of face recognition.The fractal theory can be used for object recognition,because the image fractal transform carries the spatial distribution of the image,and this method are invariant to rotation,scaling,translation and illumination,and are robust to the change of face expression.There are two methods of face recognition using fractal theory:One is using fractal coding method to face recognition.The other is using fractal dimension to face recognition.In this thesis,the rate of face recognition and the speed of identification is studied,and the main research contents are as follows:(1)The face recognition algorithm based on fractal coding is studied,and improved the classifier.The nature of this method is fractal compression,and this method mainly using the main characteristic of low frequency information,ignoring some of the high frequency information of the face image,so it can smooth the change of facial expression or little cover of face.In order to further cut down the high frequency component of the image,the improved algorithm use the wavelet transform to obtain the low frequency components in classifier design,therefore,the recognition rate has been further improved.(2)To further improve the speed of face recognition,using the idea of average face in the classifier,getting the average face of each category face image.Firstly,we use the fractal code of average face to get some candidate category.Then,the accurate result is obtained in these candidate categories.The speed of face recognition is accelerate to some extent.In the concrete experiment,using different distance measure to get the best face recognition rate.(3)In the process of face feature extraction,the image edge is used as the feature of the face.Fractal dimension can describe image texture,so it can be used to obtain the edge information of the image.The experimental result show that the fractal method can get more effective edge information than the sobel operator method.Using the ORL face database,the fractal method has higher recognition rate than the sobel operator method.(4)Using multifractal method for face recognition.The multifractal analysis give full consideration to the fractal characteristics in different levels in the process of geometry formation.And in the process of facial feature extraction the multifractal singularity is used to instead the fractal dimension.In addition,the ORL face database have the change of face pose and position.In order to avoid the influence of face recognition because of the change of face position.We use the nature of Fourier transform,getting amplitude spectrum of the image,and block the image to obtain the multifractal spectrum after the image were Fourier transformed,so that the face recognition rate is increased.