Adaptive shape prior in graph cut image segmentation

This paper presents a novel method to apply shape priors adaptively in graph cut image segmentation. By incorporating shape priors adaptively, we provide a flexible way to impose the shape priors selectively at pixels where image labels are difficult to determine during the graph cut segmentation. This is in contrast to the use of shape priors indiscriminatively at all pixels in existing image segmentation approaches, which may fail if the parameters for the shape prior term are not chosen appropriately. We integrate the proposed method in two existing graph cut image segmentation algorithms, one with shape template and the other with the star shape prior. To determine the need for a shape prior at each pixel, our experiments make use of either the original image or an enhanced version of the original image by smoothing. Experimental results in multiple application domains demonstrate the generality and superior performance of our adaptive shape prior method.

► We present a novel method to apply shape priors adaptively in graph cut segmentation.
► We provide a flexible way to impose the prior selectively.
► We integrate the proposed method in two existing graph cut algorithms.
► Our experiments make use of the original image intensity to determine the need for a shape prior at each pixel.
► Experiments in multiple application domains demonstrate the generality and superior performance.