Classification of Shoeprint Images Using Directional Filterbanks

With the abundance of evidence data collected from scenes of crime (SoC), shoeprint (or soleprint) traces usually constitute one of the hardest types of evidence for a criminal to remove before leaving the SoC. Traditional approaches to shoeprint representations attempt to classify shoeprint images based on a number of possible patterns. Such approaches are difficult to implement in an automatic fashion without the intervention of a forensic specialist. In this paper, we propose a fully-automated system to assist forensic specialists in rapidly classifying a shoeprint image found in a scene of crime (SoC). The proposed multiresolution-based system uses directional filterbanks (DFBs) to capture both local and global details in a shoeprint in a compact representation called ShoeHash. Experimental results based on forensic shoeprint databases of more than 1000 images are presented to validate the effectiveness of the proposed system in extracting shoeprint features and achieving good performance