作者: Florian Ciurea , Dan Lelescu , Kartik Venkataraman , Gabriel Molina
DOI:
关键词: Confidence map 、 Pixel 、 Parallax 、 Similarity (geometry) 、 Geography 、 Parallax mapping 、 Position (vector) 、 Artificial intelligence 、 Computer vision
摘要: Systems in accordance with embodiments of the invention can perform parallax detection and correction images captured using array cameras. Due to different viewpoints cameras, results variations position objects within scene. Methods provide an accurate account pixel disparity due between cameras array, so that appropriate scene-dependent geometric shifts be applied pixels when performing super-resolution processing. In several embodiments, detecting involves competing subsets estimate depth a location image from reference viewpoint. number generating estimates considers similarity multiple spectral channels. certain confidence map indicating reliability estimates.