RaFSi - a Multi-Platform, A Fast Watershed Algorithm Based on Rainfalling Simulation

Stanislav L. Stoev

In this paper, we present a fast watershed algorithm based on the rainfalling simulation. We present the various techniques and data structures utilized in our approach. Throughout this work, the processing of large data sets (images as well as volume data) is especially emphasized. The results' correctness, the fast execution time, and the memory requirements are discussed in detail. First we introduce a sequential algorithm and discuss the cases, where the known algorithm produces erroneous results. Afterwards, the presented watershed algorithm is compared with immersion based watershed algorithms with respect to running time and memory requirements.