Parallel segmentation based on topology with the associative net model

This paper presents an implementation of a topological segmentation on a SIMD massively parallel computer based on reconfigurability and asynchronism: Associative Mesh. This architecture provides powerful computational primitives that can apply an associative operator over the connex sets of a graph. So, basic primitives combine communications and computations. These primitives can be easily and efficiently realised in hardware by means of asynchronous operations and are adapted to a large number of image analysis primitives. We try to show the adequacy of Associative Mesh computing model with the different data movements that are generated by the several approaches of the image analysis. We are interested here with a new approach: image topology. We indicate how to get an homotopic kernel and a leveling kernel with parallel algorithms. Such kernels may be seen as "ultimate" topological simplifications of an image. This kind of image is similar to a very good split because it is based on topological information of image. We show one example of merge: we implement a method segmenting without the need of defining and tuning parameters.