Illumination-insensitive image representation via synergistic weighted center-surround receptive field model and weber law

Image representation under complex illumination condition is a great challenge in computer vision field. Motivated by Weber law and synergistic center-surround receipt field model, we propose an illumination-insensitive feature descriptor, named as Weber synergistic center-surround pattern (WSCP), including three components: differential synergistic excitation pattern (DSEP), synergistic straight orientation pattern (SSOP) and synergistic diagonal orientation pattern (SDOP). To further enhance the discriminative power of WSCP, we present a scale and orientation weighted WSCP (WWSCP), which fully considers the inner and outer layer pixels’ excitation distributions, as well as their orientation information importance. To acquire more discriminative and rich features, we utilize the spatiograms to describe patterns instead of the conventional histograms, which can obtain higher order spatial information. In the final classification stage, we present a novel distance measurement model called weighted similarity measurement model (WSMM) to improve classification accuracy, by sufficiently utilizing the information contents and orientation distributions of each pattern. Extensive experimental comparisons between our methods and other state-of-the-art methods are conducted on CMUPIE, YALE B, YALE B Ext, FERET, PhoTex, Alot and RawFooT databases, and performance results have verified the effectiveness and efficiency of our proposed methods on the robustness to illumination variations.