Virtual viewpoint synthesis using CUDA acceleration

Abstract

Depth image based rendering (DIBR) is an effective method for virtual view synthesis. But synthetic images often contain ghost effect and some holes of varying sizes. To cope with the ghost effect, this paper first uses color correction of reference views, and then combines of two fusion methods which are depth-based color image fusion and directly color image fusion. In order to fill the remaining disocclusions, this paper uses the image inpainting with depth map. Because the process of image inpainting is time-consuming, this paper implements the inpainting algorithm on GPU to accelerate the inpainting process based on Compute Unified Device Architecture (CUDA). The experiment results show that our proposed method improved the quality of virtual view synthetic images, and the image inpainting algorithm based on CUDA implemented also significantly reduce the time of hole filling.