The demo contained in this package is described in 

A General Preprocessing Method for Improved Performance of Epipolar Geometry Estimation Algorithms
Maria Kushnir, Ilan Shimshoni
Submitted to PAMI 

Please refer to the paper for details


The main function in this package is:
out = main("Filename1","Filename2","Outputname",Fs_flag,run_BLOGS_flag,angles_flag,angles_list)    

Example:
main('GEO00029.jpg','GEO00038.jpg','output_29_38',1,1,1);



Input:
Filename1 - image 1 filename
Filename2 - image 2 filename
Outputname - ouput filename
Fs_flag - 1 for entire algorithm, 0 for probabilities based on 2keypoint matches only.
run_BLOGS_flag - 1 to run BLOGS after our preprocessing method, 0 otherwise.
angles_flag - 1 for relative roll angle estimation by our method, 2 for using relative roll angles given by user and 0 for non fixed SIFT direction.
angles_list - relative roll angles in radians, relevant only in the case of angles_flag=2.



Output:
For every relative roll angle i,
out(i).matches - putative feature matches between the two images.
out(i).probabilities - probabilities that each match is correct.

Incase of run_BLOGS_flag=1 BLOGS_output.m is created with
BLOGS_output.F - Fundamental matrix.
BLOGS_output.inliers - matches that are BLOGS_output.F inliers.



Note:
The demo contains several executable files.
In case of different platform one might be required to rebuild them from their original C++ sources.
To do so please run AGPM_compile.m.