Fusion and Allocation Network for Light Field Image Super-Resolution

Light field (LF) images taken by plenoptic cameras can record spatial and angular information from real-world scenes, it is beneficial to fully integrate these two pieces of improve image super-resolution (SR). However, most the existing approaches LF SR cannot fuse at levels. Moreover, performance hindered ability incorporate distinctive different views extract informative features each view. To solve core issues, we propose a fusion allocation network (LF-FANet) for SR. Specifically, have designed an operator (AFO) among views, (SFO) deep representation Following operators, further strategy propagate features. In stage, interaction block (IIFB) supplement all views. For output are allocated next AFO SFO distilling valid information. Experimental results on both synthetic datasets demonstrate that our method has achieved same as state-of-the-art methods. preserve parallax structure generate faithful details images.