A flexible architecture for multi-view 3DTV based on uncalibrated cameras

Real-time virtual view synthesis using light field

“…The holes [2,3], artifacts [4,5], deformation [6], and abnormal edge [7] will appear after image warping; repair works for these issues, such as inpainting, have high computational complexity. Although there are a lot of recent work on improving imaging quality in real time [8], Ho [9] and Xiao [10] reduced the computational complexity and improved operational efficiency in their work but still cannot achieve real-time performance. Mori [11] and Zinger [12] enhanced imaging quality in their work but further increased computational complexity.…”

Kinect-Variety Fusion: A Novel Hybrid Approach for Artifacts-Free 3DTV Content Generation

“…In this paper we present a novel hybrid parameterized variety-based approach that uses only sparse depth information from Kinect [11] to efficiently represent the scene, and provide a low cost, feasible solution to render high quality images for multi-view 3D displays. Unlike depth image-based rendering (DIBR) algorithms [4], [16], this integrated vision-sensor technique does not require dense, accurate, complete per-pixel depth information of each color image. Instead, it recovers sparse depth information of few reference scene points and render consistent novel perspective images using only a small number of sample views of a scene without explicit 3D reconstruction.…”

“…Instead, it recovers sparse depth information of few reference scene points and render consistent novel perspective images using only a small number of sample views of a scene without explicit 3D reconstruction. Until now, DIBR techniques are mainly adopted by 3DTV systems for maintaining moderate data size during 3D content production [4], [16]. With improved active depth sensing technology, even the heavy computation involves in scene geometry extraction can be avoided.…”

“…Despite that very less work has been noticed where direct application of ranging cameras with DIBR can be made to generate novel content for 3DTV displays [19], [20]. This is mainly because the visual quality of synthesized views using DIBR highly depends on the precision of depth map quality [4]. However, the Kinect raw depth data directly cannot be useful.…”

“…The restored depth maps are not completely accurate. These depth inaccuracies are the source of severe rendering artifacts during view synthesis using DIBR [4], [16], [19], [20]. However, the presented variety-based approach implicitly use only sparse depth information of scene points and parameterize the complete 3D scene space.…”