Edge Detection Based on Primary Visual Pathway

Abstract

Edge is an important feature and edge detection matters a lot to image processing and pattern recognition. In consideration of high properties of human visual system in image perception, a brain-inspired edge detection model based on primary visual pathway was raised. According to the mechanism of lateral geniculate nucleus (LGN) cells and simple cells in primary visual cortex (V1), first, the difference of Gaussian function was adopted to model the concentric receptive field (RF) of a LGN cell; then, cell groups were created by the union of LGN cells with same property; next, RFs of simple cells with a certain preferred orientation were created by combining those of cell groups; finally, the responses of all V1 simple cells were gained by integrating the responses of different simple cells. The proposed model reflects orientation selectivity of simple cells. Results on the USF database indicate that it has better anti-noise performance and robustness in edge detection compared with the previous methods. It effectively combines brain-inspired intelligence with computer vision.