Spatial to temporal conversion of images using a pulse-coupled neural network
Ethan L. G. Brown, Bogdan M. Wilamowski · 2003
An electronic model of a pulse-coupled neural network is proposed. The model exhibits very interesting features such as segmentation, feature extraction, orientation independence and noise tolerance. Segmentation means that the output pattern depends strongly on the spatial location of the pixels in respect to one other. Feature extraction means that if the input image includes several patterns, then it is very likely the temporal output is a superposition of features in that image. The output temporal pattern is independent of the orientation of image or orientation of fragments of the image. With relatively low noise (less than 10%) the output pattern is virtually independent of the noise.