A Fast Segmentation Algorithm Revisited

Image and, more recently, video segmentation form an important part of image understanding. Several algorithms presuppose that this step is efficiently and correctly completed by an oracle. A recent bottom up image segmentation algorithm incorporates both local and global information in an image and runs fast in practice. This algorithm is loosely based on Kruskal’s algorithm on minimum spanning trees. We build on this work and use the classical Prim’s algorithm which is a theoretically superior algorithm. Our experimental result indicates that the quality of our segmentation does not suffer, yet the speed does improve. We also prove our algorithm is good in the same sense as the original work.