Fingerprint image segmentation based on rectangular block

The state-of-the-art algorithms of fingerprint segmentation, usually based on square block, are too dependent on the images of high quality and regular shape, so they have many deficiencies in dealing with low quality or irregular fingerprint images, such as high complexity, time-consuming and unsatisfactory segmentation results, etc. In order to adapt to different quality images, the proposed algorithm follows the natural characteristics of human fingers and implements an adaptive segmentation based on rectangle block. Firstly, the images are divided into non-overlapping rectangular blocks with rows and columns of the ratio of 4:3. Then, the algorithm, according to the statistical analysis, will determine whether each block is the prospect or not and end with the removal of isolated blocks by a smoothing filter. Experimental results show that the proposed algorithm has the advantages of time-saving and self-adaptability in dealing with different quality and shape of fingerprint images in comparison with the baseline.