Face Photo-Sketch Recognition using Local and Global Texture Descriptors

The automated matching of mug-shot photographs with sketches drawn using eyewitness descriptions of criminals is a problem that has received much attention in recent years. However, most algorithms have been evaluated either on small datasets or using sketches that closely resemble the corresponding photos. In this paper, a method which extracts Multi-scale Local Binary Pattern (MLBP) descriptors from overlapping patches of log-Gabor-filtered images is used to obtain cross-modality templates for each photo and sketch. The Spearman Rank-Order Correlation Coefficient (SROCC) is then used for template matching. Log-Gabor filtering and MLBP provide global and local texture information, respectively, whose combination is shown to be beneficial for face photo-sketch recognition. Experimental results with a large database show that the proposed approach outperforms state-of-the-art methods, with a Rank-1 retrieval rate of 81.4%. Fusion with the intra-modality approach Eigenpatches improves the Rank-1 rate to 85.5%.