Denoising And Inpainting Algorithem For Finger Vein Image Based On Gabor Texture Feature

The performance of finger vein recognition system is very dependent on the quality of the collected image.However,the factors that various kinds of noises produced by the acquisition equipment during imaging and transmission,as well as the dirt on the mirror surface of the equipment in the open use scene and the molting of the user's fingers,which will have a great impact on image quality,increase the difficulty of subsequent feature extraction,and ultimately affect the recognition performance of the whole system.For the problem that the existing image denoising algorithms and inpainting algorithms do not accurately use the texture feature information of finger vein image during processing,which lead to the blurring of vein texture edge after denoising,even the loss of some vein information,and the rupture of vein edge after inpainting,a finger vein image denoising and inpainting algorithm based on Gabor texture features is proposed.While eliminating noise or inpainting damaged areas,making full use of the texture feature information of the image to protect the texture edge structure of vein better,which has good reference for vein image processing.The specific research contents are as follows:1.Vertical phase difference coding method for texture features of finger vein image based on Gabor wavelet is proposed.Compared with the gray structure and gradient information used in traditional image denoising algorithms or inpainting algorithms,Gabor wavelet has better scale and direction performance to match the characteristics of different finger vein thickness and extension direction,and has certain robustness to light intensity change and noise interference,which make it more suitable to extract texture edge information of finger vein image.Gabor texture features obtained by coding can describe vein texture trend more accurately.Theoretical analysis and experiments prove that the Gabor texture feature has excellent local feature ability and good discrimination,and can accurately describe vein texture edge information,which can be used in the subsequent research of finger vein image denoising and inpainting algorithm.2.The NLM denoising algorithm of finger vein image based on Gabor texture feature and BC dual kernel function is proposed,which has two innovations: First,the concept of Gabor texture similarity is proposed,and then the similarity between image blocks is measured by texture similarity and gray value Euclidean distance to better remove the impact of noise,which solves the problem that the traditional NLM algorithm adopts the gray value Euclidean distance alone,and the similarity calculation accuracy is easily affected by the light intensity and noise.The second is to propose the Butterworth Cosine dual kernel function,which is more in line with the requirements ofthe ideal kernel function of the NLM algorithm,and has better retention of the vein edge,solves the problem that the traditional NLM algorithm's exponential kernel function is insufficiently weighted for the image block with high similarity and the image block with low similarity is excessively weighted,which easily leads to the blurring of the vein edge detail.Theoretical and experimental analysis prove that compared with the traditional NLM algorithm and improved NLM algorithm,the algorithm has better ability to protect the texture edge and detail information of vein while denoising,and the quality of finger vein image after denoising is higher.3.A finger vein image inpainting algorithms based on Gabor texture constraints is proposed.According to the characteristics of local texture coherence in finger vein image,the algorithm uses Gabor texture constraint mechanism to avoid the influence of introducing too much information with low texture correlation on the restoration area,which makes the vein texture of the repaired image more coherent,improves the performance of subsequent feature extraction and matching recognition.The theory and experiment prove that the Gabor texture constraint makes the repaired finger vein image by this algorithm have better visual coherence,improves the quality of the image,the feature of vein skeleton extracted by subsequent segmentation is more stable,and ultimately improves the recognition performance of the damaged image,compared with the traditional inpainting algorithms which do not use the texture information of finger vein image accurately.