Biologically-inspired model for multi-order coloring texture boundary detection

It presents a hybrid-level texture image boundary detection algorithm inspired from human visual system (HVS). The proposed algorithm integrates three important visual primitives: luminance, texture, and color into a functional system. The paper focuses on relevant fundamental researches on HVS and systematic integration to investigate the task of texture boundary detection thoroughly. It employs the encoding form in HVS with systematic integration to build up a complete algorithm for texture boundary detection. Color images are firstly decomposed into three opponent axes and the 1(st)-and 2(nd)-order features are extracted by a Gaussian filter and Gabor filters. With the proposed adaptive weights selecting mechanism, the hybrid-order boundary can be obtained. Among extensive tests, boundaries between uniform textures can be detected successfully and accurately. For textures that are non-uniform or non-regular, the results also reflect some meaningful properties which are consistent to human visual sensation. In addition to satisfying testing results, processing employed in this algorithm is very simple and intuitive with only few assumptions and no training procedure involved. Compared with the present researches, the proposed algorithm has a good application potential.