Pulse coupled neural networks with adaptive synapses for imagesegmentation
Jörg Schreiter , Ulrich Ramacher , Arne Heittmann , Daniel Matolin and René Schüffny
A network of integrate-and-fire neurons with reciprocal synaptic connections to the four next neighbors is considered. The input to each neuron is the feature of an image, i.e. light intensity. Like in the biological archetype synchronization shall be used as an indicator if the input in certain neurons are related or not. Two rules for the unsupervised adaption of the synaptic weigths have been derived to achieve segmentation image of areas belonging to the same feature.
Full Text: PDF