An Hierarchical Labeling Technique for Interactive Computation of Watersheds

The watershed computation is a prevalent task in the geographical information systems. It is used, among other purposes, to forecast the pollutant concentration and its impact on the water quality. The algorithm to compute the watershed can be hard to parallelize and with the increasingly data growth, the need for parallel computation increases. In this paper we propose a new method to parallelize the watershed computation. Our algorithm is decomposed into two tasks, the parallel watershed segmentation into a hierarchy that allows in a second task to retrieve randomly large watersheds at run-time in interactive time.