In this paper we propose a method for approximately minimizing objective functions of discrete functions with continuous value domain. Whereas graph-cut originally addresses binary labeling problems, there exist several algorithms which use graph-cut to find local minima of multiple-label problems. Many practical problem in the area of computer vision are even continuous-valued, and optimization methods of graph-cut type cannot be applied directly.
This is different with the proposed method, which is based on a two-step processing of the data. In the first step, binary graph-cut is used to generate regions of support within different ranges of the signal. In a second step, a robust error minimization is approximated based on the previously determined regions. The proposed method compares favorably to multi-label graph-cut and robust smoothing for the application of disparity estimation w.r.t. quality of results and computational speed.
@inproceedings{f07a,
Author = {Felsberg, M.},
Booktitle = {SSBA},
Date-Added = {2007-08-07 11:14:41 +0200},
Date-Modified = {2007-08-07 11:14:41 +0200},
Title = {Extending Graph-Cut to Continuous Value Domain Minimization},
Year = {2007}}