Error Detection in Panoramic Videos: a Pairwise Assessment within Stitching

Sandra Nabil 1 Frédéric Devernay 1 James L. Crowley 2
1 IMAGINE - Intuitive Modeling and Animation for Interactive Graphics & Narrative Environments
Inria Grenoble - Rhône-Alpes, LJK - Laboratoire Jean Kuntzmann, INPG - Institut National Polytechnique de Grenoble
Inria Grenoble - Rhône-Alpes, LIG - Laboratoire d'Informatique de Grenoble
Abstract : One way to provide realistic immersive VR content relies on producing high-quality panoramic videos. These videos are usually produced using multiple cameras with different optical centers and which may not be perfectly synchronized This results in spatial and temporal artifact, even though the blending algorithm strives to reduce them. In this paper, we devise a method that detects potential visual artifacts, based on existing view synthesis quality metrics. The method works by computing pair-wise quality at each blending step and fusing them to produce a global map of potential errors. To get a more accurate prediction, we develop a mask that is then applied to the error map and therefore accentuates the defects on the blending cutting line. Results show that the calculated distortion map succeeds to identify visual artifacts in panoramas which can help design better solutions to this problem in the future.
Complete list of metadatas

Cited literature [14 references]  Display  Hide  Download
Contributor : Sandra Nabil <>
Submitted on : Thursday, July 26, 2018 - 12:56:01 PM
Last modification on : Thursday, May 2, 2019 - 3:30:31 PM
Long-term archiving on : Saturday, October 27, 2018 - 12:14:48 PM


Files produced by the author(s)


  • HAL Id : hal-01849267, version 1


Sandra Nabil, Frédéric Devernay, James L. Crowley. Error Detection in Panoramic Videos: a Pairwise Assessment within Stitching. 2017. ⟨hal-01849267⟩



Record views


Files downloads