Checking the literature about image segmentation, if I am not wrong, there are two main approaches:
- Variational approach e.g. level-set optimization in Chan-Vese
- Combinatorial approach e.g. using graph cut in GrabCut
Can someone give me some high-level comparison between these two directions?
For example, I feel with level-set optimization one can easily incorporate image structure (e.g. edges) into the energy function, which is not easy in Grabcut. Am I correct?