Catalogue of Artificial Intelligence Techniques
Aliases: Region Growing
Categories: Pattern Recognition and Image Processing
Author(s): Bob Beattie
The basic idea of region finding is to produce a segmentation of the image in which the regions (connected sets of pixels) found have a useful correspondence to projections of scene entities such as objects or surfaces. As such it is the (currently unpopular) dual of Edge Detection/Boundary Detection. There are two main approaches:
- Start with the maximum number of regions (e.g., make every pixel a region) and merge adjacent regions based on some measure of similarity until a satisfactory segmentation has been achieved.
- Start with a few (possibly one) large regions and recursively split them into smaller regions based on some measure of dissimilarity until a satisfactory segmentation has been achieved.
- Zucker, S.W., Region growing: childhood and adolescence Computer Graphics and Image Processing 5 (1976), 382--399.