Human vision system is a very complicate information processing system, which behaves as bottom-up or top-down closed loop control system. Retinal level, thalami and visual cortical level, and thinking level are three levels in human vision system, correspondingly there exist three levels in visual information processing, (1) pixel, (2) geometric shape (texture or color region), and (3) object based on knowledge. The three vision levels work together, no primary and secondary, a complete vision process can be activated by any level of them. Motivated by these considerations, and in order to simulate human vision information processing process, we corresponding design three experiments of our visual computing model based on the visual cortical neural mechanism, (1) image segmentation based on perceptual uniformity clustering (pixel level), (2) vision recognition based on deformable template (region level), and (3) shape extraction based on knowledgebase (knowledge level).
展开▼