Improved view synthesis with depth reliability maps

In this paper, an improved view interpolation algorithm with depth reliability maps is presented to enhance the quality of synthesized images. First, a region under the mask within a virtual view image is treated as a set of indicating pixels. Next, based on depth reliability maps, a segmentation mask can be produced by subdividing the set into three subsets, corresponding to areas which an intermediate image should be blended from. Finally, a novel view image along the baseline axis can be generated with the weighted interpolation of the predict images based on the mask. Experimental results demonstrate that the proposed view interpolation scheme provides some improvements in subjective visual perception and objective measurement in terms of peak signal to noise ratio and structural similarity index over the traditional methods.