Source code for MRF minimizationHere is C++ source code for MRF minimization and all benchmarks. The code in the first 3 files compiles with gcc 4 under linux and cygwin on 64-bit machines. 64-bit code is are assumed by default; edit the Makefile for 32-bit machines.
The code is provided "as is" as a service to the community. It comes with NO WARRANTIES WHATSOEVER, and we cannot provide any help on how to use it beyond the documentation included here.
- the MRF library
mrfstereo.zip (352K, contents, README.txt) - a stereo matcher frontend calling the MRF libary
imageLib.zip (339K, contents, README.txt) - a simple image libary (used by mrfstereo)
If you have an MRF energy function to minimize, you just need the MRF library.
- If you want a stand-alone stereo matcher, you need the first three files.
- To run all the benchmarks, use the last file. However, the files in MRF-benchmarks.zip are slightly older versions that have not been updated to compile on 64-bit machines.
- Some of the files refer to code on Vladimir Kolmogorov's webpage, which has moved to here. In particular, the file typeTruncatedQuadratic2D.h is now available here (download "implementation").
See the README files within each zip file for more information.
You might also be interested in the OpenGM library and/or the FastPD optimization software by Paragios and Komodakis.