Plenoptic 2.0 Intra Coding Using Imaging Principle

Xin Jin,Fan Jiang,Lingjun Li,Tingting Zhong

DOI: https://doi.org/10.1109/tbc.2021.3108058

IF: 4.5

2022-01-01

IEEE Transactions on Broadcasting

Abstract:Plenoptic 2.0 videos that are captured by focused plenoptic cameras outperform the traditional plenoptic videos with higher spatial resolution of rendered subaperture images and more effective light field sampling. However, its distinct imaging principle causes different pixel intensity distribution and complex macropixel structures, which results in ineffectiveness in coding them using the existing plenoptic video compression methods. Thus, a novel compression method is proposed in this paper to dramatically improve the intra prediction efficiency for multi-focus plenoptic 2.0 video coding. It consists of two new intra prediction modes: imaging-principle-guided static prediction (ISP) and imaging-principle-guided zoomed prediction (IZP). In ISP, the prediction candidates spatially positioned by plenoptic 2.0 imaging principle are weighted by minimizing boundary error to generate the predicted samples. In IZP, the spatial positions and scales of the prediction candidates vary according to the focusing status and the focal length of the microlens that generates the pixels in the prediction candidates. The combination of two modes generalizes plenoptic 2.0 video coding for multi-focus plenoptic cameras. The proposed approach outperforms high efficiency video coding (HEVC) standard by an average of 15.37% bitrate reduction under All Intra configuration. It also outperforms other state-of-the-art methods with obvious compression efficiency improvements.