Skip to Main Content
We present an algorithm and its implementation on a GPU for the view synthesis of 3D video objects. The algorithm is based on the theory of trifocal transfer using dense depth information of two cameras. We present issues that arise when using the pipeline architecture of the GPU for non-computer graphics algorithms and introduce approaches how to solve these problems with focus on our application. In order to assess the quality achieved by the algorithm, we present a method for generating ground truth both for synthetic and real imagery. We present first results of subjective tests for quality evaluation.