Data driven Pre attentive Texture Segmentation by using Markov Random Fields

Abstract

One of the properties of the human visual system is its pre-attentive texture discrimination capability. The human visual system can perceive textual differences almost instantaneously and effortlessly. It has been hypothesized that pre-attentive texture discrimination are based on a few local conspicuous features(textons). We propose a coupled Markov random field with two layers as a computational model for the texton-based pre-attentive texture segmentation. The model prameters are estimated and the segmentation is carried out iteratively.