Velocity tuned filters and motion perception

Image processing, with developing technology, has been a very important filed for research. Being relatively new topics, video processing and detennining of optical flow are interesting topics, because there are plenty of applications of them. For neural networks that can mimic some characterics of biological livings and especially their implementations on analog VLSI are real-time, they have very fast operation rates and they are, with this property, superior than dijital processors. In this thesis, optical flow, afterwards velocity tuned filters implemented with asymmetrical diffusion networks are studied. Velocity tuned filters, inspired by retinal structure in the vertebrates are realised with active resistive diffusion networks. Afterwards, an analog network sufficient for motion detection is studied from photoreceptor level to velocity estimation. Velocity tuned filters yields efiicient results despite its simplicity. Velocity tuned filters can be realised easily with CNN circuits or as an integrated circuit analog VLSI technology.Keywords: Velocity tuned filters, retina, perception of motion, optical flow.