Inpainting a 3-D scene using stereo extraction

摘要

The present invention relates to systems and methods for rendering three-dimensional scenes with improved visual characteristics from a pair of two-dimensional (2D) images having different perspectives. The 3D scene obtains a first two-dimensional (2D) image of the scene object from a first viewpoint, obtains a second 2D image of the scene object from a second viewpoint different from the first viewpoint, and a first 2D image And generating a depth map from a second 2D image, generating a depth map and a 3D scene from the first 2D image and a second 2D image, detecting regions of an initial 3D scene having incomplete image information, and It is created by reconstructing the detected areas of, determining replacement information, modifying the reconstructed areas, and rendering a 3D scene with reconstructed areas modified from multiple perspectives.