Coarse-to-fine least squares stereo matching for 3-D reconstruction

Least squares methods have been used for 3-D reconstruction from digital stereo images. A preliminary estimate of the disparity of points to be matched in the two images is required for the convergence of the method. A new approach based on a Gaussian pyramid decomposition of the stereo images and application of the least squares method at each level of the pyramid is presented. Results obtained from digitised stereo images and 3-D reconstruction are reported.