Take Your Model Further: A General Post-refinement Network for Light Field Disparity Estimation via BadPix Correction

Most existing light field (LF) disparity estimation algorithms focus on handling occlusion, texture-less or other areas that harm LF structure to improve accuracy, while ignoring potential modeling ideas. In this paper, we propose a novel idea called Bad Pixel (BadPix) correction for method modeling, then implement general post-refinement network estimation: Bad-pixel Correction Network (BpCNet). Given an initial map generated by specific algorithm, assume all BadPixs it are in small range. Then BpCNet is modeled as fine-grained search strategy, and more accurate result can be obtained evaluating the consistency of images limited Due assumption between input output, perform network, work almost iteratively. We demonstrate feasibility our theory through extensive experiments, achieve remarkable performance HCI 4D Light Field Benchmark.