Context

Light field coding has been extensively researched in the last few years. A wide range of light field coding solutions have been proposed, from well-known, off-the-shelf standard codecs, to those specially designed for light field data. These methods aim to exploit the huge amount of redundant information about the light rays in the same scene that convey angular and spatial information. To achieve compression performance, a light field compression method needs to characterize the relationship among a number of viewpoints, as well as the spatial correlation within the views. Learning-based approaches present advantages to solve complex tasks.

The scope of the JPEG Pleno Light field learning-based coding activity is the creation of a learning-based coding standard for light field targeting competitive compression efficiency compared to state-of-the-art light field coding solutions and supporting a royalty-free baseline. To this end, the purpose of this workshop is to exchange experiences, to present technological advances and research results on learning-based coding solutions for light field data.

Practical information

• Date: June 22, 2022
• Time: from 13:00 UTC to 15:00 UTC
• Access: Zoom

Program

• 13:00-13:10 UTC - Welcome & Introduction (Carla Pagliari, Peter Schelkens and Saeed Mahmoudpour)
• 13:10-13:30 UTC - Learning-based lossy and lossless compression of light fields (Giuseppe Valenzise, Université Paris-Saclay, CNRS, CentraleSupélec)
• 13:30-13:50 UTC - Deep Decoding of Light Field Images (Søren Forchhammer, DTU Electro, Technical University of Denmark)
• 13:50-14:10 UTC - Learned plenoptic image compression with spatial-angular decorrelation (Xin Jin, Shenzhen International Graduate School, Tsinghua University)
• 14:10-15:00 UTC - Questions & Discussion

More information

Interested parties are invited to join the JPEG Pleno Light Field mailing list and regularly consult the JPEG.org website for the latest news.