Video stitching with spatial-temporal content-preserving warping

Wei Jiang, Jinwei Gu

Abstract: We propose a novel algorithm for stitching multiple synchronized video streams into a single panoramic video with spatial-temporal content-preserving warping. Compared to image stitching, video stitching faces several new challenges including temporal coherence, dominate foreground objects moving across views, and camera jittering. To overcome these issues, the proposed algorithm draws upon ideas from recent local warping methods in image stitching and video stabilization. For video frame alignment, we propose spatial-temporal local warping, which locally aligns frames from different videos while maintaining the temporal consistency. For aligned video frame composition, we find stitching seams with 3D graphcut on overlapped spatial-temporal volumes, where the 3D graph is weighted with object and motion saliency to reduce stitching artifacts. Experimental results show the advantages of the proposed algorithm over several state-of-the-art alternatives, especially in challenging conditions.