Translation invariance-based super resolution method for mixed resolution multiview video

Mixed resolution format plays an important role in 3DTV coding scheme. In recent years, several research works discussed the Super Resolution (SR) problem in Mixed resolution format, where the SR image is created using the up-sampled low resolution (LR) image and the high frequency components borrowed from the adjacent high resolution (HR) warping image. However, the quality of the obtained HR images is impaired by the misalignment error caused by depth error in 3D warping. A novel SR reconstruction model is proposed in this paper. It is composed of three components: structure term, detail information term and the regularization term. They were used to preserve the structure similarity of the LR image, to preserve the detail information from the adjacent HR image, and to ensure uniqueness of the solution respectively. Simulation results show that the proposed method achieves high performance in PSNR.