Palm Vein Image Enhancement Algorithm Based On Image Fusion And Deep Learning

With the development of the Internet,the connection between people has become more interconnected,identity authentication as the premise of information security has become particularly important.In view of the security problem of traditional identity recognition,biometric identification has developed rapidly because of its uniqueness,stability and difficulty in forgery.As a typical biological feature,the palm vein is protected by the skin of the palm,so it is difficult to photograph and image under visible light;the palm is often in a semi-clenched state,which is difficult to steal;as a 'living feature',it can only be certified if the blood is in flow.Based on the above characteristics,the palm vein has become the most promising biologic feature.A large number of palm vein features recognition technology has developed rapidly.However,in the process of collecting palm vein image,its imaging effect is sensitive to skin scattering,light change,device performance,etc.Yet obtaining high quality palm vein image is the basis to obtain high recognition rate.Therefore,the preprocessing of palm vein image is a very important step in the palm vein recognition system.On the basis of observing a large number of palm vein images,and studying in depth about the reasons which affect the quality of palm vein images,it is found that except for the grayscale inconsistency of palm vein image caused by uneven illumination and the noise of palm vein image caused by skin scattering,palm print imaging differences also affect the image quality of the palm vein.Based on the above discussion,this paper presents two effective methods to improve the image quality of palm veins:1.The method of palm vein image enhancement based on image fusion:using the traditional preprocessing method to process the image to obtain the initial enhancement image,and according to the pixel size and the grayscale value of palm print in the palm vein image,eight directional filters for palm print extraction are proposed to extract palm print structure and weighted fusion between palm print extraction image and the initial enhancement image.The experimental results show that the proposed method can effectively remove image noise and palm print interference,enhance vein texture and improve the contrast between vein information and background information,and it can be better applied to three kinds palm vein recognition system.2.The method of palm vein image enhancement based on deep learning:Aiming at the shortcomings of the method of palm vein image enhancement based on image fusion,such as sensitive to abnormal data,long processing time,and there are some surface texture and noise in the palm vein image.By analyzing the conditional Generative Adversarial Network,the paper propose to enhance the palm vein images by using conditional confrontation-generating networks.The paper uses the palm vein enhancement images obtain by the method based on image fusion as a condition against the network target image,making the original palm vein image learning the vein network structure of the target image,while "discarding" a small amount of the palm surface texture imaging and noise introduced by the enhancement method based on image fusion,to achieve the purpose of enhancing vein texture.The experimental results show that the deep learning method proposed in this paper can effectively remove the a small amount of palm print and noise in the palm vein enhanced image obtained by the image fusion-based enhancement method,thereby effectively enhancing vein texture.