Automatic shoeprint classification based on DFT

Shoeprints are important forensic evidences found at crime scenes. Quick retrieval and classification method are needed since the shoeprint image databases are very large. Two main categories of shoeprint images are stripes and concentric circles. Since these kinds of images have apparent periodicity which can be easily detected in frequency domain, in this paper we propose a DFT-based classification method. The proposed method is much faster than Hough transform based method since the parameters search scopes are significantly decreased in frequency domain. Detailed procedures of the proposed method and experimental results are given in the paper.