Relative pose estimation algorithms, such as 5-point algorithms, often return multiple solution candidates. Random sample consensus (RANSAC) algorithm can be used to find the the correct solution from among the solution hypotheses and remove incorrectly matched feature points.

Here you can find the Matlab implementation of the QuEst 5-point algorithm incorporated with RANSAC. Download all files and run Example.m to see how a unique relative pose estimate can be obtained from two images. A C++ implementation of QuEst (without RANSAC) is available at https://github.com/KavehFathian/QuEst

All scripts are written and tested in MATLAB R2017b, 64-bit Windows OS. MATLAB’s Computer Vision System Toolbox and Statistics and Machine Learning Toolbox are required. See the Readme file for more details.