Generating Robust and Stable Disparity Map Using Phase-Based Correspondence Matching from Stereo Video Sequence

Abstract

This paper proposes a robust and stable disparity map generation algorithm using phase-based correspondence matching. When considering the application of auto stereoscopic displays, an accurate, robust and stable disparity map generation algorithm is indispensable to synthesize multiview images from a stereo image pair. The proposed algorithm considers a motion between the previous and current frames estimated by phase-based correspondence matching to generate the accurate and stable disparity map in the temporal axis. Through a set of experiments using still stereo images and stereo video sequences, we demonstrate that the proposed algorithm exhibits efficient performance of disparity map generation compared with the conventional algorithms.