NATURAL VIEW TRANSITION IN THE PANORAMA BASED NAVIGATION USING FAST FOURIER TRANSFORM TECHNIQUES

摘要

This paper proposes a new algorithm that is able to smoothly and realistically transit the scene from one viewpoint to another in the panorama based navigation system. The proposed algorithm is divided into two steps: one is the prewarping that aligns the viewing direction in two panorama images, and the other is the bidirectional disparity morphing(BDM) that generates the intermediate scene with the prewarped panorama images. In the prewarping, we first analyze the phase relation of two images using the fast Fourier transform (FFT), and then compute the scale, the rotational movement, and the translation between two images. These information make two images to be aligned. In the BDM, we estimate the block based disparity vector (DV) and use two occluding patterns in order to correct the DV for the occluded block. We make an experiment on the proposed algorithm with the panorama images and as a result, the user can move naturally from one panorama image to another.