Image Semantics Segmentation using Watershed Algorithm

In this paper a novel image semantics segmentation algorithm is proposed, which combines edge and region-merged based techniques. First, an edge-preserving statistical noise reduction approach is used as a preprocessing stage in order to compute an accurate estimate of an image gradient. Second, we segment image into primitive regions by applying watershed algorithm on the image gradient magnitude. The watersheds computation algorithm used is based on immersion simulations, that is, on the step of the recursive detection and fast labeling of the different catchment basins using queues. At the end, we merge neighboring region into homologous region using morphological erosion and dilation. Some experiments are presented to illustrate availability and effectiveness of our approach