VVS: Versatile View Synthesizer for 6-DoF Immersive Video

Autor: Jung, Joël, Boissonade, Patrick
Přispěvatelé: Orange Labs, Jung, Joel
Jazyk: angličtina
Rok vydání: 2020
Předmět:
Popis: Immersive video services require the transmission of multiple views captured by a convergent camera array and/or by 360 omni-directional cameras. Unfortunately, to enable a six degrees of freedom (6-Dof) navigation in a scene, displaying captured views is not sufficient. A significant amount of intermediate views have to be synthesized, to render the pixels that perfectly match the motion of the user. This is typically achieved by depth image based rendering. This paper presents the Versatile View Synthesizer, adopted as the reference MPEG-I Visual algorithm for 6-DoF activity. Following strict common test conditions, experimental results report a significant 250% BD-rate saving, corresponding to a 1.4dB improvement compared to the previous VSRS4.2 reference algorithm. Those objective results are confirmed by MS-SSIM and VMAF metrics, and by subjective evaluations applied on pre-defined navigation paths.
Databáze: OpenAIRE