A Markov random fields model for hybrid edge- and region-based color image segmentation

A framework based on a Markov random field approach for color image segmentation enhanced by edge detection is presented. We use a previously developed methodology to transform the image into an R´G´B´ space to remove any highlight components preserving the vector-angle component, representing color hue but not intensity, to remove shading effects. To improve the segmentation process we describe the idea of a line process. This allows for the integration of region segmentation with edge detection in a Markov random field framework. We discuss the advantages of this new model with respect to the previously developed image segmentation model.