Computer texture boundary detection based on human visual perception model

Texture boundaries detection is considered a difficult problem. In human visual perception, however, discriminating textures is an "effortless" process. Based on the human pre-attentive texture perception model called texton, we propose a set of computational algorithms to extract textural features and form the texton energy map. We then use a neural positive feedback algorithm for boundary extraction. The model and algorithms are tested on simulated and real-life pictures and are found to yield very satisfactory results.<<ETX>>