The technology herein describes a method of converting video from its standard “stereo” format to “multiview” content suitable to be used to create 3D effects. The purpose of this conversion is to allow for any media to be viewed on automultiscopic displays in 3D without the need for the glasses required by Stereoscopic 3D methods. The conversion is facilitated by a combination of two rendering approaches, the more traditional Lagrangian approach, and the more recent Eulerian approach.
The Lagrangian method focuses on the depth and disparities in the input images. This provides the advantage of being able to account for large disparities, as well as perform non-trivial disparity manipulations via extrapolation and interpolation. In contrast, the Eulerian method is aimed at processing local changes in pixel values. It has the advantages of high performance, subpixel accuracy, and sensitivity to ambiguous depth cases. Through these two techniques, the video input is decomposed, disparity information is calculated for each of the resultant component, and the output is reconstructed according to the disparities, to produce the desired 3D effect.