A new paradigm for segmentation

The major problem in the segmentation task is a problem of decision. The authors propose five guidelines to solve the problem and present the software architecture of a low-level segmentation system that would solve the problem according to the guidelines. They have already developed one module: a region growing method that aggregates pixels according to the type of the region and some thresholds calculated dynamically. Some results obtained are very encouraging. The authors attempt to detect edges with the same philosophy