Iterative computation of camera paths in an image based rendering application

Abstract

This thesis presents a novel algorithm to iteratively compute camera paths of long image sequences. Scale Invariant Features are first extracted from the ordered set of images. These images are then matched pair-wise sequentially and correspondences are computed. An initial geometric path is found after by applying a bundle adjustment algorithm on these correspondences. Distances between cameras are computed from this initial estimation. The iteration process starts by grouping nearby cameras and then bundle adjusting the groups, and ends by merging the groups. This process is repeated until the reprojection errors fall into the preset tolerance. The key point in this algorithm is to take advantage of loopbacks in the image sequences. We have obtained excellent results for two camera paths, namely the spiral path and the snake like path. Our algorithm achieves both precise and stable results.