Novel Algorithm For Finger-Vein Image Enhancement And Identification

Biometric identification technology is a new kind of authentication technology in the 21st century,which has good research value and broad application fields.Among them,the hand vein recognition has great market potential due to its advantages of non-touch,difficult to imitate,vivo identification,high safety,simple and easy to use,and has drawn great attention from various research groups and industry at home and abroad.However,in practical applications,finger vein identification technology also encountered some bottlenecks,including:(1)The quality of the finger vein image depends largely on the quality of the acquisition device,so the design of the acquisition device has a certain requirements;(2)Due to the light intensity in the collection process,the position of the finger is not fixed and other factors,the original image is noise or offset,resulting in large differences in sample images of the same individual;(3)Finger vein feature information is not rich,compared with the facial features,its characteristics are not obvious;In practice,low-quality finger vein images are often encountered,which may affect the feature extraction and identification of vein features.Therefore,the enhancement of details of vein image texture is an important study in finger vein recognition technology.In this paper,the weighted guided filtering enhancement algorithm of finger vein image based on edge detection,the image quality evaluation algorithm based on improved structural similarity and the extraction algorithm of vein image feature based on improved Weber local descriptors are studied.The main research contents are as follows:(1)The weighted guided filtering enhancement algorithm of finger vein image based on edge detectionGuided filters are mostly used in image denoising and defogging,in this paper,the original guided filter is applied to the enhancement algorithm of finger vein image,the experimental results are not satisfactory,the original guided filter theory is improved by comparative experiments and theoretical analysis.The weighted guided filter of edge detection is proposed that through calculating the amplitude value of the vein image edge operator to punish the fixed neat parameters in the guided filter,adaptive neat parameters of the vein texture area and the smooth area are obtained respectively,which makes the filter possess better edge protection feature.The proposed weighted guided filtering enhancement algorithm of finger vein image based on edge detection can retain and highlight more vein texture details of the image while keeping the noise reduction effect,and the enhanced image is conducive to the subsequent features extraction and recognition.(2)Improved Structural Similarity(SSIM)algorithmIn addition to enhancing the effectiveness of the algorithm by identifying experiments,it is also possible to test the algorithm through subjective and objective image quality assessment methods.The former is difficult to implement in practice.Therefore,the latter is the hot spot in the field.The research on the objective image quality evaluation method is to make its quality evaluation model reflect the subjective evaluation more accurately.Zhou Wang et al,proposed an objective method of image quality evaluation:structural similarity(SSIM)Based on the fact that human visual system(HVS)can extract structural information from images or videos,a large number of experiments show that SSIM's image quality evaluation algorithm performs better than MSE and PSNR.Due to the different information contained in different regions of the image,the human eye can only focus on the information of interest at the same time.Therefore,this paper proposes a SWSSIM image quality evaluation method based on edge detection weighting.The enhanced image plus Window to find the window edge detection amplitude and SSIM index,and then summed to obtain the image SWSSIM index.(3)The extraction algorithm of vein image feature based on improved Weber local descriptorsThe original WLD theory is mostly used in the field of face recognition.In this paper,WLD is applied to the feature extraction algorithm of the finger vein image.The experimental results are not satisfactory.Through comparative experiments and theoretical analysis,it is found that the original WLD differential excitation adopts the method of difference summation,The noise is too sensitive and there is a problem that the positive and negative differences cancel each other out.As a result,the extracted feature can not effectively represent the characteristic detail information of the finger vein image,that is the edge information.In order to solve this problem,we use the gradient magnitude based on the edge detection as the difference threshold to calculate the differential excitation,and propose a Weber local descriptor(EDGWLD)based on the edge detection gradient magnitude.Firstly,utilizing the edge detection algorithm to extract the edge information of the image;Secondly,the gradient amplitude is optimized according to the position of the pixels in the image,and the discrimination is improved by increasing the directivity of the gradient amplitude.According to Weber's law,make the ratio of the optimized gradient amplitude and the current pixel gray value as a differential excitation;Finally,the improved differential excitation and direction are fused to obtain EDGWLD.