Segmenting Non-random Noise Biomedical Images

This paper presents a new approach for 2D object segmentations using an automatic method applied on images with severe non random noise conditions. Results from biomedical images (cytologies) are presented. The proposed methodology works from the output of an edge detector, which is processed to obtain an approximation of the shape of the object from partial information. The nal estimation of the shapes is obtained tting a stochastic deformable template model.