This software implements the binary energy minimization method presented in
[1] "Superdifferential Cuts for Binary Energies"
Tatsunori Taniai, Yasuyuki Matsushita, and Takeshi Naemura
In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2015)
pp.2030--2038, Boston MA, US, June 2015.
If you use this software for research purposes, you should cite the aforementioned paper in any resulting publication.
You need to additionally download the following data and code:
-
GrabCut Dataset
Available at http://research.microsoft.com/en-us/um/cambridge/projects/visionimagevideoediting/segmentation/grabcut.htm -
MATLAB Wrapper of the Boykov-Kolmogorov max-flow min-cut algorithm
Availabe at http://vision.csd.uwo.ca/code/
Modify the file paths in Run_Benchmark.m and run it. Note that this code may produce slightly different results from the ones originally reported in [1], where another C++ implementation was used.
Currently the code only implements the histogram matching segmentation using L1 and L2-distance (Type-1). Algorithms for Type-2 (fractional higher-order energies) and Type-3 (pairwise non-submodular energies) will be implemented in the future.
- Released the MATLAB implementation for the Type-1 algorithm.
Copyright (c) Tatsunori Taniai All Rights Reserved.