A novel pixon-representation for image segmentation based on Markov random field

In this paper, a pixon-based image representation is proposed, which is a set of disjoint regions with variable shape and size, named pixon. These pixons combined with their attributes and adjacencies construct a graph, which represents the observed image. A Markov random field (MRF) model-based image segmentation approach using pixon-representation is then proposed. Compared with previous work on region-based and pixon-based segmentation methods, the present method has some remarkable improvements over them. Firstly, a set of significant attributes of pixons and edges are introduced into the pixon-representation. These attributes are integrated into the MRF model and the Bayesian framework to obtain a weighted pixon-based algorithm. Secondly, a criterion of GOOD pixon-representation is presented and a fast QuadTree combination (FQTC) algorithm is proposed to extract the good pixon-representation. The experimental results demonstrate that our pixon-based algorithm performs fairly well while reduces the computational cost sharply compared with the pixel-based method.