Walking to singular points of fingerprints

• We propose a novel fast method for singular point detection.
• We can walk directly to singular points without scanning fingerprint images.
• Simplicity of the idea makes the algorithm easy to be implemented.
• Our algorithm can be combined with state-of-the-art method to improve performance.

Singular point is an essential global feature in fingerprint images. Existing methods for singular points’ detection generally visit each pixel or each small image block to determine the singular point. That is to say, existing methods require scanning the image to compute a quantity at each pixel or block, and hence they are inevitably time-consuming. We propose a fast algorithm for detecting singular points by walking directly to them instead of scanning the image. Walking Directional Fields (WDFs) are established from the orientation field. Then following the walking directions on WDFs, we can rapidly walk to the singular points. The walking algorithm is extremely fast and easily implemented with acceptable accuracy. Further more, its accuracy can also be improved by combining with state-of-the-art methods: we can rapidly walk to a candidate singular point, then refine its location using existing more accurate method in the local area. Experimental results on datasets of SPD2010 and FVC validate the high efficiency and satisfactory accuracy of the proposed algorithm.