![]() Our approach leads to greater temporal coherence and yields a more consistent segmentation across views, compared to existing techniques. Our technique is based on the computation of an affinity measure between feature tracks, designed to process stereo videos. In this thesis, we present a novel sparse stereo-video segmentation technique that builds on existing 2D sparse segmentation frameworks. Current state-of-the-art techniques cannot replace semi-automatic rotoscoping, but they alleviate the work of artists by limiting the amount of required interaction. The far reaching goal is that these segments correspond to semantically meaningful video objects. Our work exploits stereo-disparity information from a stereo scene in conjunction with long-term trajectories to improve on state-of-the-art approaches.Īutomatic video segmentation techniques are aimed at creating masks to delineate coherent regions in a video. We investigate the adaptation and extension of existing techniques for monoscopic videos so as to process stereoscopic sequences. Both algorithms need temporal stability as well as view consistency. ![]() In this thesis, we study two diffcult fundamental aspects of stereo-video post-production, for which existing tools are far from matching the abilities of artists: segmentation and inpainting. The fexibility offered by digital video is a key factor to the recent success of stereoscopic cinema, as most artifacts can be corrected or attenuated during post-production so as to deliver a pleasant 3D experience throughout. Recent developments of digital stereoscopic videography call for new tools to help carry out post-production tasks for this medium. ![]() It has found numerous applications from cinema post-production to television sets. Nowadays, video processing technology is ubiquitous.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |