We are pleased to announce that our paper “Steered Mixture-of-Experts for Light Field Video Coding” has been accepted. It will be published in the proceedings of SPIE Optical Engineering and Applications (Applications of Digital Image Processing XLI). The paper has been presented in the SPIE optics + photonics conference, as part of the session on plenoptic imaging.
Steered Mixture-of-Experts (SMoE) is a novel framework which is able to represent multidimensional image modalities.
In this paper, we propose a coding methodology for SMoE models that is readily extendable to any dimensional model. SMoE is inherently multi-dimensional, and thus can represent any image modality of any dimension.
Firstly, we present the coding approach of SMoE models for light field video and we evaluate the coding performance. Light field video is a 5D image modality consisting of time, two angular, and two spatial dimensions.
Furthermore, we compare the performance of three multi-view HEVC (MV-HEVC) configurations that differ in terms of random access. We specifically interpret each sub-aperture view from the light field video as a single view in MV-HEVC.
Finally, we discuss the results which validate that SMoE offers excellent coding performance. When compared to MV-HEVC for low- to mid-range bitrates, SMoE achieves savings up to 75%.
Comparison between MV-HEVC and SMoE for coding light field video at low bitrates.
More details can be found in the proceedings: https://www.spiedigitallibrary.org/conference-proceedings-of-spie/10752.toc#PlenopticImaging