Image segmentation using watersheds guided by edge tracing
Lei Ma, Jennie Si, Glen P. Abousleman · 2002
We present two image segmentation algorithms. The proposed algorithms are the integrated watershed and adjacent region merging (IWARM) algorithm and the self-guided watershed and ARM (SGWARM) algorithm. Due to the integration of noise reduction and watersheds, IWARM saves up to 25 percent processing time over existing leading algorithms such as seeded region merging (SRM) algorithm and region adjacency graph (RAG) algorithm on a variety of test images. The IWARM algorithm also provides better segmentation accuracy over the two mentioned algorithms. SGWARM was developed to improve processing speed by eliminating any "out-of-interest" regions before they are provided to IWARM for further processing. Both algorithms have been tested extensively and are shown to provide excellent performance using aerial and MRI imagery.