Texture segmentation with a cascade correlation neural network using markov random fields

摘要

In this paper a Cascade Correlation Neural Network is used to segment an image according to some known textures. The texture labeling process is modeled using a Markov Random Field (MRF); this model allows to classify a pixel in a determined texture using information from its neighbors. The neural network learns the characteristics of the MRF that model the texture labeling process for a set of known textures.