Research On Pretreatment Algorithm For Hand Geometry Image

With the wide applications of information and networks techniques, personal authentication has become one of important methods for security of information and networks. Because of the disadvantages of traditional personal authentication methods, biometrics identification techniques appear. Because of the simple collection instrument and the high verification speed, hand geometry identification has been another biometrics identification method except fingerprint. The pretreatment part of hand geometry image is mainly studied in this thesis.The pretreatment hand geometry image is the base of hand geometry recognition system and the essentially important step for hand geometry character extraction and identification. In the hand geometry image processing, some important information of image, such as edge and detail character, are often submerged by noise signal, and huge effects on subsequent processing of image, such as edge detection, image segmentation and image match, are produced. So proper pretreatment for noisy image is very important in the image processing and has significance for image quality improvement.Some familiar image pretreatment denoising methods are introduced in the thesis. The average filtering denoising method, wavelet threshold denoising method, extremum median filtering combine with wavelet transformation denoising method and wavelet maximum value denoising method are reseached chiefly. Then the thesis analyses many kinds of edge detection arithmetic operators and the Palm shape image pretreatment is simulated in MATLAB.For wavelet threshold denoising method, the thesis through many times experiments sets suitable threshold level and selects estimated wavelet coefficient to get better wavelet threshold denoising effect. The essence of adapting wavelet maximum value denoising method is researched deeply in the thesis. At the question of setting threshold, it adopts adaptive threshold level to overcome the question which single threshold level can't make the largest separation between signal and noise in every scale. The method put forward in this paper not only cancels noise effectually but also keeps edge details of the image. It has good effects of canceling noise.In the research of median filter and wavelet transformation, the thesis proposes extremum median filtering combine with wavelet transformation denoising method. The method combines with two method's advantages, cancels image noise effectively, it has better effects than wavelet threshold denoising method or extremum median filtering.From simulation result of some pretreatment methods, we can know the adapting wavelet maximum value denoising method adapts palm shape image more than others.