Ghost Elimination via Multi-Component Collaboration for Unmanned Aerial Vehicle Remote Sensing Image Stitching

Ghosts are a common phenomenon widely present in unmanned aerial vehicle (UAV) remote sensing image stitching that seriously affect the naturalness of results. In order to effectively remove ghosts and produce visually natural results, we propose novel method can identify eliminate through multi-component collaboration without object distortion, segmentation or repetition. Specifically, our main contributions as follows: first, ghost identification component locate potential area; detect significantly moving objects two stitched images. particular, due characteristics UAV shooting, images small quality is poor. We mesh-based difference comparison ghosts; use an tracking algorithm accurately correspond each pair. Second, design information source selection strategy generate replacement region, which replace located avoid Third, find process elimination mosaic by eliminating caused initial blending with selected source. validate proposed on VIVID data set compare Homo, ELA, SPW APAP using peak signal noise ratio (PSNR) evaluation indicator.