Image Segmentation For Fingerprint Recognition

Abstract

Fingerprints have been widely used for personal recognition in many forensic applications. The segmentation of fingerprint images is a fundamental step in recognition systems. It classifies pixels of the image into two classes the foreground and the background. This paper proposes an improved method for fingerprint segmentation using a histogram-based thresholding approach. The main idea is to model the fingerprint images as a Gamma distribution conversely to the current approaches that uses Gaussian distribution. In fact, in some type of images, the image data does not best fit in a Gaussian distribution, but rather a Gamma distribution. Choosing the suitable distribution means enhancing the quality of the segmentation. The proposed method was applied on bi-modal fingerprint images and promising experimental results were obtained. It was implemented and tested on a large set of real fingerprint images. A comparative study between our Gamma based approach and the traditional Gaussian approach based on some quantitative measures showed that our technique yields better performance results and high quality segmented images. This means that the segmentation based on a Gamma based approach is better than a Gaussian based approach for Fingerprint images.