Parameterized Feasible Boundaries in Gradient Vector Fields

aries as a model-based segmentation procedure. This enSegmentation of (noisy) images containing a complex ensemhances the quality of segmentation, as not only local image ble of objects is difficult to achieve on the basis of local image information is used, but also the global shape of the object information only. It is advantageous to attack the problem as captured in the model. This can be used to guarantee of object boundary extraction by a model-based segmentation closed boundaries even if for some part of the boundary procedure. Segmentation is achieved by tuning the parameters the image information does not indicate an edge. of the geometrical model in such a way that the boundary As, in general, the object shape is not fixed, we need a template locates and describes the object in the image in an model which is parametrically deformable. Model-based optimal way. The optimality of the solution is based on an segmentation is achieved by tuning the parameters of the objective function taking into account image information as geometrical model in such a way that the boundary temwell as the shape of the template. Objective functions in literaplate locates and describes the object in the image in optiture are mainly based on the gradient magnitude and a measure describing the smoothness of the template. In this contribution, mal way. The optimality of the solution is based on an we propose a new image objective function based on directional objective function taking into account image information gradient information derived from Gaussian smoothed derivaas well as the shape of the template. In the special case tives of the image data. The proposed method is designed to where the object shape is fixed and known, model-based accurately locate an object boundary even in the case of a segmentation reduces to standard template matching. We conflicting object positioned close to the object of interest. We consider the general case here. further introduce a new smoothness objective to ensure the Apart from yielding a closed boundary, parameterized physical feasibility of the contour. The method is evaluated on deformable templates have the important property that artificial data. Results on real medical images show that the features of the object boundary (like its curvature) can be method is very effective in accurately locating object boundaries in very complex images.  1996 Academic Press, Inc. computed analytically from the template parameters. This reduces the considerable loss of shape information caused by the fact that edges are confined to lie on a discrete