The images observed on an automultiscopic screen correspond to a cut through the light field created by the screen. Unlike in real world, the light fields produced by automultiscopic screens have a repetitive structure. When the viewing position is not optimal this induces image quality-degrading visual artifacts. To avoid transitions artifacts, the light field produced by an automultiscopic display must be continuous. This method employs techniques in seamless image and video editing to cut and stitch together different copies of the light field images into one natural, continuous composition.
Local and global shearing techniques are applied to refine the light field by aligning the structure of its repetitive fragments. These fragments are then overlapped to find an optimal cut, upon which they are stitched together using Poisson reconstruction techniques. This method produces reconstructed, continuous light fields with reduced disparities and transition artifacts, therefore exhibiting improved visual quality. In contrast to previous hardware solutions, this optimization does not require knowledge about viewer’s position, which makes the technique suitable for an arbitrary number of observers.