Image Segmentation with a Statistical Appearance Model and a Generic Mumford-Shah Inspired Outside Model

摘要: We present a novel statistical-model-based segmentation algorithm that addresses recurrent problem in appearance model fitting and model-based segmentation: the "shrinking problem". When statistical models are fitted to an image order segment object, they have tendency not cover full leaving gap between real detected boundary. This is due fact cost function for evaluated only on inside of object at boundary detected. The state-of-the-art approach overcome this shrinking detect edges force adhere these edges. Here, we introduce region-based motivated by Mumford-Shah functional does require detection In addition model, define generic estimated from input outside model. Shrinking prevented because misaligned would create large discrepancy inside/outside method independent dimensionality image. apply it 3-dimensional CT images.