A recursive low level vision system

The paper addresses a very important, yet one of the difficult issues in computer vision and visualization-low level vision modeling. It proposes a novel low level vision model which recursively integrates adaptive filtering, segmentation and edge detection. The model has strong biological merits: a) the model architecture is based on a biologically inspired neural network-network of networks which simulates human visual cortex; b) evolutionary computation is applied to identify the hierarchy and clusters in the network. But the model does not constrain itself by the biological facts. Instead, it proposes that by using clustering method, adaptive filtering, segmentation and edge detection are naturally linked to one another. The feasibility of the concept is demonstrated via a visual example.