3D content generation for autostereoscopic displays

Abstract

Content generation is one of the most critical issues for the growth of 3DTV services in future. New autostereoscopic displays, such as the Philips Wow® 3D display have significant advantages, including improved 3D viewing experience, wider viewing angles, multiple viewers and no need for special glasses. However, due to their content formatting requirements (2D + depth), live-action content is much more difficult to create. In this paper a new approach for 3D content generation is proposed, by integrating an existing state-of-the-art MRF-based disparity estimation method with additional pre- and post-processing steps. The proposed method uses rectification and colour segmentation to solve significant problems in disparity estimation as well as decomposition of the scene in foreground and background depth maps. Tests with different stereo sequences have already produced very promising results.