A Connected Component Approach to the Watershed Segmentation

The watershed transformation is a popular image segmentation algorithm for grey scale images. In this paper, we present a deenition of the watershed segmentation, based on sets of neighboring pixels. The deenition is consistent with the behavior of most implementations of the watershed algorithm which do not construct watersheds lines, namely, to chose one arbitrary label in the case of competing labels. Moreover, diierent distance metrics to approximate the geodesic distance on plateaus can be incorporated into the deenition. The relation to the traditional dee-nition of the watershed segmentation is proven in the paper. The deenition leads to a new type of watershed algorithm which is closely related to the connected component algorithm. Our complexity analysis shows that for practical cases, the algorithm has a linear complexity with the number of pixels, independent on the image content and number of grey-levels. The algorithm is simpler with respect to implementation and data structures. Additionally, the memory requirement is small and independent of the number of grey-levels in the input image. Furthermore, our timing results show a signiicant improvement in the running time, compared against the classical watershed algorithm based on hierarchical queues.