Fusion of Intra- and Inter-modality Algorithms for Face-Sketch Recognition

 Combination of Eigentransformation, a global intra-modality method [1], with the Eigenpatches local intra-modality approach that has been used for face hallucination [2]. For both approaches, photos are transformed to sketches or vice versa to reduce the modality gap. This allows recognition using a traditional face recogniser.  These algorithms are further fused with an inter-modality method called Histogram of Averaged Orientation Gradients (HAOG) [3]. The Chi-Square (χ2) histogram matching method is used to evaluate the distance between the HAOG descriptors of probe sketches and those of the gallery photos.  The scores output from each face recogniser are normalized using min-max normalisation in the Normalisation module and are finally fused with the Fusion module using sum-ofscores method. The algorithms are evaluated on the Chinese University of Hong Kong Face Sketch FERET (CUFSF) database [4]. This database contains viewed hand-drawn sketches of subjects in the Color FERET database [5]. A total of 842 subjects were considered.