Abstract: The fingerprint image in preprocessing is often affected by many factors, but sometimes can not meet the requirements of fingerprint identification system. In this article, an effective fingerprint preprocessing algorithm is presented based on the traditional methods for preprocessing. Firstly, a method based on block gradient variance is implemented to achieve the separation of the fingerprint image and the background area. Secondly, according to the fingerprint feature, we use orientation map and mean filter to enhance the image. And using simplified Gabor filter improves the filtering template to filter edge blur effect. After binarization, thinning and eliminating pseudo minutiae, we get the fingerprint image skeletonization and extract the fingerprint feature points. Experiments show that the preprocessing algorithm is applicable to images of different qualities and levels. It is flexible, efficient, easy, and accurate, meets the requirement of fingerprint identification system.

Key words: fingerprint identification, block gradient variance, eliminating pseudo minutiae, mean filtering, binarization