The proposed work involves a hybrid data representation using efficient light field coding. The existing light field coding solution are implemented using sub-aperture or micro-images. However, the full capacity in terms of intrinsic redundancy in light field images is not completely explored. This paper represents a hybrid data representation which explores four major redundancy types. Using coding block, the most predominant redundancy is exploited to find the optimum coding solution that provides maximum flexibility. To show how efficient the hybrid representation works, we have proposed a combination of pseudo-video sequence coding approach with pixel prediction methods. The observed experimental results shows a positive bit rate saving when compared to other similar methods. Similarly, the proposed method is also said to outperform other coding algorithms such as WaSP and MuLE when compared on a HEVC-based benchmark.
Light Field Image Coding Using Dual Discriminator Generative Adversarial Network And VVC Temporal Scalability
