Two View Reconstruction

Author:Jin Zhou (ferryzhou@gmail.com)

Overview

Do 3D reconstruction from two images

Usage

tvr version 0.1:

usage: tvr -L leftImage -R rightImage -o Xs.txt -c cams.txt [options]

reads two image file

runs two view reconstruction

write 3D points to Xs.txt and cameras to cams.txt


options:
-a alpha       the ratio between focal length and image width (default 1.0)
-s             sift method (default)
-k             use pyramid LK method
-p param       corner detector params ('haris 400 0.02 8'(default) or 'grid 16')
-f threshold   threshold of normalized ransac fundmamental matrix computation (default is 0.015)
-r threshold   threshold of reprojection error (default is 2.0)
-d             draw intermediate images
-w             write intermediate images
                           one view correspondences image to cos1.jpg
                           two view correspondences images to cos2.jpg
                           epipolar lines to eplines1.jpg and eplines2.jpg
                           back projection of inliers images to bx1s.jpg and bx2s.jpg
-u             extension name of images, default is jpg
-P prefix      prefix of images
-m             output 3d mesh model to model.off and model.vrml
-t filename    output point correspondences file
-q             quiet (turn off debugging output)

Examples

example1: read left image and right image, do reconstruction, show and write intermediate images, output 3d model:

tvr -L left.bmp -R right.bmp -o Xs.txt -c cams.txt -m -d -w