Fingerprint Image Denoising Via the Improved Total Variation (TV) Algorithm

摘要

This paper proposes several improvements for a fingerprint image denoising method that is based on nonlocal total variation (TV) models using split Bregman iteration. The main improvement involves the addition of relaxation factors to the two-step iterative process of split Bregman iterative algorithms to obtain a double relaxation split Bregman iterative algorithm. The improved method is tested using numerous fingerprint images from FVC2004 databases. The experimental results show that the improved double relaxation split Bregman iterative algorithm achieves significantly better performance in terms of the visual subjective evaluation and the quantitative objective evaluation. The method achieves better noise suppression and effectively retains image edge details.