4 - Modélisation et mise en correspondance avec la pyramide neuractive

This paper introduces the neuractive pyramids to model objects under elastic deformations . One pyramid is built on each frame of the image sequence . Each level of the pyramid is a regular graph that is recursively built on a low-pass version of the original picture . Cells of this graph deform to model the local information of the picture . Deformations are obtained by the minimization of an energy function computed both on the gradient of the picture and the graph structure . Each cell contains a vector of statistical moments computed on its domain and its neighborood . Matching pyramids defines a vector Field of local elastic transformations . The matching operator is based on a self-organizing map, introduced by Kohonen . Softness and multiresolution aspects of the pyramids allow accurate and robust results . An application to the matching of 2D cardiac MRI scans shows the interest of the method for deformable objects .