Panorama construction using binary trees

Abstract

We present a novel algorithm for computing a panorama from a video sequence, which is based on AVL binary trees and is called PCBT. The main characteristic of PCBT is to compute automatically the reference frame, which is used to project the rest of frames on a planar projection. We use the homography as the camera motion model, it is computed only frame to frame, and the rest of the homographies needed for compositing the panorama are automatically computed through the binary tree using the previously computed homographies. The experiments show that PCBT improves the results of Autopano Giga and Image Composite Editor, two well-known applications, and the results of Stitching_Detailed sample of the OpenCV library, for computing panoramas under a planar projection.