Novel optical flow optimization using pulse-coupled neural network and smallest univalue segment assimilating nucleus

Optical flow, with abundant local motion information, has been widely investigated in the last few decades. To improve the robustness and accuracy of optical flow estimation, we proposed a systematic optimization algorithm based on the pulse-coupled neural network (PCNN) and the smallest univalue segment assimilating nucleus (SUSAN). The primary aim of the proposed algorithm is to overcome the problems incurred by the noise disturbance and texture insufficiency. Specifically, our method performs a homogeneous area extraction based on texture variation using a nonlinear filter, smallest univalue segment assimilating nucleus. Then a novel 3-D pulse- coupled neural network model is constructed to perform optical flow optimization. Because of the excellent clustering capability of the PCNN, the proposed algorithm significantly improves the quality of optical flow estimation in the presence of noise. The enhanced optical flow field and extraction results are combined to solve the problem of insufficient texture. The proposed algorithm is evaluated in both synthetic and real testing images to demonstrate its excellent performance.