Research On Image Enhancement Of Finger Vein

In the information age,Biometric recognition have been concerned by different social fields and widely used in various domains of safety with the strong demand for the reliability and efficiency of identity verification.The technology of biometrics recognition could identify person by using powerful computer based on the biologic feature or behavior feature of human being.The typical features include face,fingerprints,palm print,iris and retina.Depending on its particular advantages of high-security,liveness detection and user-friendly,finger vein recognition technology has attracted the interest of researchers at home and abroad.Besides,the finger vein image capturing equipment is so small that we can promote the marketization of this technology easily.Nevertheless,it is difficult to obtain high quality images actually because of the surrounding environment and subjects' behaviors.For example,the various performances of devices,lighting variance,thicknesses of muscles,the position of finger and so on.And that leads to degradation of image.With Low quality images comes the issue of background noise,low contrast and luminance variance,which reduces distinguishability of veins.Stable feature often could not be extracted if we directly carry on segment on these low quality images.Worse still,it has a negative impact on the performance of finger vein recognition.Furthermore,low quality images add to complexity of the follow-up algorithms Therefore,how to effectively enhance the finger vein image is also a key issue for finger vein recognition.We propose a new adaptive enhancement framework based on finger vein image quality evaluation.First of all,in order to obtain the quality of images,we train a new classifier based random forest model to classify the image as high quality or low quality.Next,we respectively use simple enhancement for high ones and complex enhancement for low ones.Experimental results show that the proposed finger vein image quality evaluation is feasible and the adaptive frame of enhancement overcomes their disadvantages of single method in terms of time and performance.Regarding to the fact that the vein details are hardly to be completely reserved,and the contrast of the enhanced image is not high enough.We apply the enhancement method using sparse representation on finger vein image.Two corresponding dictionaries are generated.One is the greyscale dictionary generated from the original finger vein images,the other one is the binary dictionary from the corresponding segmented images.The patches of dictionary are selected through adjusting parameter.Then the input target image is represented by greyscale dictionary to get the sparse coefficients via a sparse coding process.Finally,the enhanced blood vessel image is obtained from the solved sparse coefficients and binary dictionary.Experimental results show that quality of finger vein image is improved.