Automatic Latent Fingerprint Segmentation based on Orientation and Frequency Features

Latent Fingerprint contains as small regions of fingerprint. The small number of minutiae and the noise characteristic of latents make it difficult to automatically match latents to their mated full prints that are stored in databases. To overcome this problem, automatic segmentation is used. An important step in an automatic fingerprint recognition system is the process of segmenting the fingerprint images. In this paper, the segmentation process consists of the three blocks namely (i) Orientation feature, (ii) Frequency feature and (iii) Post processing. The Orientation Feature process is used to obtain the fingerprint ridge orientations, the Frequency Feature process is used to obtain the local ridge frequencies of the fingerprint. The Post processing method combines these two results to find the candidate fingerprint regions. The proposed scheme achieves 77% accurate segmentation for latent matching of fingerprint. This system improves the accuracy segmentation of foreground and background regions from latent fingerprint images.