Finger Vein Image Processing Method Research Combining Finger Crease

With the development of science and technology, now information grows exponentially, so information security is gradually getting more and more attention. Biometric recognition technology is inherent, unique, safe and convenient, so it begins to dominate authentication field instead of traditional authentication mode, such as keys, cards and so on. Owing to the advantages of high recognition rate, anti-counterfeiting and low cost, finger vein recognition is gradually becoming a new favorite in biometric recognition field.Because finger vein images captured by existing finger vein image acquisition technology contain a lot of noise and have low contrast, finger vein images should be preprocessed to make finger veins more clear before the finger vein recognition. Based on the image database of Hong Kong polytechnic university, the paper did a deep research on the finger vein image preprocessing, and the main work included the following several aspects:Firstly, finger vein images and corresponding finger texture images in finger image database were subjected to preprocessing steps which automatically removed the areas with much noise and little information and kept the areas with key information. Then the background information of finger vein images was removed and the region of interest was kept by image processing algorithms, such as Canny edge detection operator, binarization algorithm and so on.Secondly, to overcome rotation interference in finger vein image acquisition process, we calculated the direction of the finger by fitting the finger outline with a linear fitting method and corrected the deflection angle of the finger.Thirdly, this paper analysized classical image enhancement algorithms and local histogram equalization algorithms, and did lots of experiments with these algorithms. What is more, an improved adaptive histogram equalization algorithm was proposed and used to enhance finger vein image, which was effective and had low time complexity.Fourthly, the finger vein texture information in enhanced finger vein images was extracted by Gabor filters and was processed by binarization algorithm. Then by using morphological algorithm to remove the noise in images and to bridge breaking portion of finger vein images, more complete finger vein texture was gotten. Finally, finger vein topology was gotten by extracting the skeleton of finger vein with morphological thinning algorithm and deburring.