Local feature fitting active contour for segmenting vessels in angiograms

An active contour model for vascular segmentation has been proposed, by defining a new, local, feature fitting, energy function. A vesselness filter is applied to the image in a directional Hessian-based framework. The filter output, as a feature, expresses the degree of the correspondence of each pixel to the vessel structure. By using intensity information obtained from local regions, the proposed model is able to solve the problem of intensity inhomogeneity in images. In addition, by introducing this feature into the fitting process, the model exhibits greater accuracy when compared to existing models. Experimental results from synthetic images and coronary X-ray angiograms verify the desirable performance of the proposed model.