60 likes | 208 Views
Parallelization of RHSEG. - Mahmood Vinayak. Problem description. Background: Image segmentation is the partitioning of an image into related sections or regions. Segmentation is a key first step for a number of approaches to image analysis and compression. Segmentation:
E N D
Parallelization of RHSEG - Mahmood Vinayak
Problem description Background: Image segmentation is the partitioning of an image into related sections or regions. Segmentation is a key first step for a number of approaches to image analysis and compression. Segmentation: A segmentation hierarchy is a set of several image segmentations of the same image at different levels of detail in which the segmentations at coarser levels of detail can be produced from simple merges of regions at finer levels of detail. Why? Change/simply the representation of an image to more meaningful format easier to analyze
Methods • HSWO: Hierarchical step-wise optimization • Region growing segmentation • Compares(merge) only spatially adjacent regions HSWO • HSEG: Hierarchical segmentation • Augmented HSWO • Merges spatially non-adjacent regions • Computational intensive HSEG • RHSEG: Recursive Hierarchical segmentation • Augmented HSEG • Merges spatially non-adjacent regions • Reasonable number of comparisons (merges) RHSEG
Parallelization Number of processors could be multiple of four, say p = 4m Partition the input image into p partitions and send each partition to the corresponding processor. If the number of recursive levels is Lr, then the recursive levels from Lr to (Lr – m) are done in all processors in parallel (each processor use his local partition of image to apply the recursive operation from level Lr to Lr-m). After that each group of four processors send their results to the leader processor of that group (processor with id is lowest). This manner continues until reaching to level 1 with only one processor.