Disparity compensation using geometric transforms

This paper describes a method to enhance the disparity compensation in 3D-stereo image and video compression algorithms. Disparity compensation is usually performed using a block matching technique between both views, disregarding the various levels of disparity present for objects at different depths in the scene. An alternative coding scheme is proposed, taking advantage of the cameras setup information and the object´s depth in the scene, to compensate more complex spatial distortions introduced by the use of convergent cameras. In order to perform a more accurate compensation, the list of reference frames is enriched with additional geometrically transformed images for the most relevant levels of depth in the scene, resulting from projections of one view to another. This scheme can be implemented in any state-of-the-art video codec, as H.264/AVC or HEVC, in order to improve the disparity matching accuracy between views. The experimental results, using HEVC algorithm, show proposed method coding efficiency, both for stereo image and video, presenting bitrate savings up to 4.87% and 2.87%, respectively.