Binary segmentation (GrabCut) - Person

Image size: 600 x 450, labels: 2
Data term: based on Gaussian mixture color models of foreground and background
Smoothness term: Potts model modulated by local contrast

Since this is a binary labeling problem, both graph-cut methods (swap and expansion) find the global optimum in a single iteration.

Spreadsheet

Program log

Input problem:

Result images:

ICM BP-S BP-M Swap Expansion TRW-S input image