Repository Moved
This project has been merged into Loopy Belief Propagation for Stereo Matching.
The new repository contains the latest code, documentation, and updates.
A C++ implementation of Loopy Belief Propagation for stereo matching that offers the choice between six message update schedules. It uses the "Min-Sum" version of the algorithm with a small improvement for better results. The improvement is that in the calculation of belief, the data cost does not add up (normally it had to add up one time).
The Tsukuba stereo image that used as input.
The disparity map that created at the output using the "Synchronous" message update schedule.
The disparity map that created at the output using the "Bipartite" message update schedule.
The disparity map that created at the output using the "Accelerated" message update schedule.
The disparity map that created at the output using the "Accelerated2" message update schedule.
The disparity map that created at the output using the "ByRow" message update schedule.
The disparity map that created at the output using the "ByRow2" message update schedule.
- Belief Propagation for Stereo Matching (Synchronous)
- Belief Propagation for Stereo Matching (Bipartite)
- Belief Propagation for Stereo Matching (Accelerated)
- Belief Propagation for Stereo Matching (Accelerated2)
- Belief Propagation for Stereo Matching (ByRow)
- Belief Propagation for Stereo Matching (ByRow2)