Solving nonuniform problems on SIMD computers: Case study on region growing

A Data Parallel Algorithm for Solving the Region Growing Problem on the Connection Machine

“…While previous parallel implementations 13,14] of the split and merge approach have used dynamic or tree structures to represent the regions in the image, our implementations use only one and two-dimensional arrays. We also introduce an element of randomness to the algorithm whenever a tie occurs during merging; this has signi cantly reduced the execution time.…”

Real time VLSI implementation of a fast split and merge segmentation algorithm

“…Initially it was proposed to select the neighbor with lowest ID to be merged among all the eligible neighbors [4]. Copty et al proposes a random neighbor selection to break the tie between multiple neighbors satisfying Best Merge Criterion (BMC) [5].…”

A Parallel Version for the Propagation Algorithm

“…The graph boundaries are the image regions and the connections between the extremities stand for the neighbors relation of the regions. The first parallel versions of the regions growing algorithm based on the "Split and Merge" approach were implemented over SIMD machines and dynamic structures were used to store image regions information [7,8]. Another experimental study of the parallel versions of the image segmentation algorithm based on the regions growing technique (also based on the "Split and Merge" approach) was presented by [9].…”