Research Of Pretreatment Algorithm On Fingerprint Image

With the development of society,the traditional safety system based on keepsake and password has been weaker,which can not meet the requirement of modern safety system.In this case,the need of a more reliable safety system with higher anti-fake performance prompts the appearance of fingerprint identification technique. This technique,with a higher safety and reliable performance,can improve the reliability of identity resolution in virtue of human body's physiological feature,and it has been a research focus these days.Fingerprint identification falls into three parts,they are pretreatment,feature extraction,and characteristic matching.The thesis mainly focused on pretreatment and feature extraction,it optimized and innovated some key algorithms of these two parts,which can be described as follows.Firstly,the algorithm dismembered and normalized the original input image, extracting the effective coverage and decreasing the variations in gray-values between ridges and valleys,that facilitates the subsequent steps.Secondly,in the enhancement part,the thesis made an improvement of traditional Gabor Filter.While computing the orientation of fingerprint,the new algorithm developed a method to evaluate the orientation of current block with consideration of its surrounding blocks, avoiding deviation of current block caused by a rapid change of its ridge orientation. While extracting the frequency of fingerprint,a step of judging the false wave crest has been added.Thirdly,in the thinning part,the thesis analyzed the cause accounts for inaccurate thinning image with many burrs and incomplete thinning in the bifurcate points,making some improvement of OPTA thinning Algorithm.It added four elimination templates and removed three reservation templates, experiment results proved a good thinning effect.Fourthly,In the feature extraction part,the thesis developed a new eight-neighborhood coding ridge tracing algorithm, removing some templates of consecutive points and bifurcate points which have been optimized and removed in the thinning algorithm,and finally marking terminate points as well as bifurcate points to execute feature extraction.Experiment result indicated that such new algorithm has a less operation but with a higher accuracy.All the algorithms introduced above have been implemented on Visual C++6.0, and result proved an adaptive good effect,which facilitates the next characteristic matching process.