Edge Detection Based on Spiking Neural Networks

A bio-inspired edge detection method based on spiking neural networks is proposed.Firstly, the image pixels are transferred into spiking trains.The first firing time of the spiking trains are used to encode the image pixels.Then, the coding in a window are inject into a neuron.If the corresponding pixels of the window are in a flat region, the coding will go to the neuron in a same time and force the neuron to produce a sparse spike trains;If the corresponding pixels of the window are in edge region, the coding will go to the neuron in different time and force the neuron to produce a dense spike trains.The edge pixel is determined by whether the firing density of the corresponding neuron is beyond a threshold.The experimental results show that the method is better than many traditional methods in mathematical sense, and is more biological realistic than the existing methods.