Integrated and hierarchical sortmap-relabel image segmentation methods

Abstract

We propose two image segmentation algorithms: the inte- grated sortmap relabel with adjacent-region merging (ISARM), and the self-guided sortmap relabel with adjacent-region merging (SGSARM). Due to the integration of noise reduction and fast merging, ISARM pro- vides a 25% improvement in processing time, as compared to leading existing algorithms such as the region adjacency graph (RAG) algorithm, on a variety of test images. ISARM also provides better segmentation accuracy than the RAG algorithm, by a measure combining the mean squared error and the number of regions obtained. SGSARM is designed for use with large images (say 102431024 or larger). It incorporates two levels of processing: an edge detection algorithm of linear complexity, which is applied to large images to detect regions of interest (ROIs), followed by ISARM for finer segmentation of each ROI. SGSARM there- fore has significant advantages in speed and accuracy when used in large images. Simulation results are provided to demonstrate the perfor- mance of both algorithms. © 2002 Society of Photo-Optical Instrumentation Engi- neers. (DOI: 10.1117/1.1511246)