Parallel watershed transformation algorithms for image segmentation

Parallel Marker-Based Image Segmentation with Watershed Transformation

“…3c). Since the emphasis in this paper is more on parallelization of the marker based region merging problem, we do not detail further the local watershed labeling (wl) (see also [18,19]). However, for the purpose of marker-based region merging, the local labeling result (o), the markers (m), the BCG (bcg), and the WNG (wng) suffice, and the explanation follows next.…”

“…The corresponding ranges of labels in the master's root array, shown in Fig. 8b, are then scattered back to processors and relabeling with the final labels is performed in every subimage (see also [18,19]). …”

Parallel image component labelling with watershed transformation

“…Thus, the total computational time is Table 1. Additional results for some artificial images can be found in [24]. Notice that the measured times do not include data loading, distribution, coalescence, and saving.…”

“…Although several trials have been previously made to parallelize watersheds (see [25], [26], [27], [28], [29]), the task is far from being easy since the operation relies on the history of region growth. Various serial methodologies for computing catchment basins with zerowidth watershed lines [7], [13], [21], [22], [23], [33] have been employed for the purpose of parallelization on MIMD computers (see [24]). In each parallel approach, the image is distributed to a virtual grid of processors and partial results are produced by any of the watershed labeling techniques referred above.…”

InterSeg: a distributed image segmentation tool

“…Image segmentation is not a new subject and a wide variety of algorithms have been proposed in the past decades (Haralick and Shapiro, 1985;Pal and Pal, 1993;Dey et al, 2010). In addition, a number of automatic tools for image segmentation are now available.…”