Hybrid Image-based Rendering for Free-view Synthesis

Image-based rendering (IBR) provides a rich toolset for free-viewpoint navigation in captured scenes. Many methods exist, usually with an emphasis either on image quality or speed. In this paper we identify common IBR artifacts and combine the strengths of different algorithms to strike good balance speed/quality tradeoff. First, address problem visible color seams that arise from blending casually-captured input images by explicitly treating view-dependent effects. Second, compensate geometric reconstruction errors refining per-view information using novel clustering filtering approach. Finally, devise practical hybrid algorithm, which locally identifies utilizes method best suited region while retaining interactive rates. We compare our against classical modern (neural) approaches indoor outdoor scenes demonstrate superiority and/or