A PCNN-FCM time series classifier for texture segmentation

Texture segmentation is a complex task in image analysis. Although many works have been done in this area, texture segmentation is still an open research area. The purpose of this paper is to investigate the potential of time signatures generated by a Pulse Coupled Neural Network, PCNN, to perform texture segmentation. Time series features are generated by the PCNN, filtered and then they are clustered by the FCM algorithm to achieved texture segmentation. A posterior morphologic process is later performed to improve the segmentation. The proposed method is evaluated against brightness, texture type and texture adjacency sensitivity. Findings indicate that the time series features capture discriminative information able to represent texture primitives. The overall performance of the proposed method on two and five texture images may indicate a promissory future for other image segmentation tasks.