Contour detection based on the interactive response and fusion model of bilateral attention pathways

Contour detection is the key pre-processing step for human vision to perceive objects. Considering the shunt transmission and interactive response characteristics of visual information in visual pathways, a new contour detection method, which is based on the interactive response and fusion model of bilateral attention pathways, is proposed. Firstly, according to the color antagonism mechanism of single-opponent receptive field (SORF) in the sub-visual cortex, a SORF dynamical adjustment model based on local luminance information was devised to realize the joint coding of luminance and color boundaries. Secondly, a multi-directional fretting method in the optimal azimuth interval was designed by simulating the direction-sensitive characteristics of the classical receptive field of the primary visual cortex. Then, a visual information interaction model based on bilateral attention pathways—which combines visual attention and bilateral shunt mechanisms of visual information—was developed to obtain spatial salient contours of dorsal attention pathway and sparse responses of ventral neurons. Finally, the multi-pathway information fusion mechanism of the high-level visual cortex was introduced, and visual information differences between bilateral attention pathways were used to obtain the target contours. The experimental results show that our method can effectively highlight contours and eliminate textures, which will provide new ideas for subsequent image understanding and analysis.