Segmentation d’images Par Les Contours Actifs

Abstract

Image segmentation by active contours is a method to locate the boundaries of object, using a curve that deforms by minimization of energies. In this memory, we propose a parallel approach for fast object tracking in medical images sequence. The proposed approach defines three Agents: The first uses a pyramidal multiresolution analysis to decompose the images of the sequence in multiple resolutions. The second agent detects the object in low–resolution images by parametric active contours. The third agent realizes the projection of the contours detected in low-resolution images to high-resolution images. Our approach was tested using three sequence types: synthetic, biological and echocardiographic, the obtained results are encouraging. Key