Systematical Diverse-Defect Kinect Depth-Map Promotion for Very Low Cost 3D Photographing

The paper proposes a structured Kinect depth-map restoration scheme to efficiently bring out visually plausible depths for visual qualified stereoscopic videos. The scheme exhibits a repairing arrangement of low-quality Kinect depth maps to obtain the very low-cost 3D photographing by efficiently integrating the well-known DSPs and our own DSPs addressed herein. Its framework consists of two repairing groups named segmentation-free discriminative filling and edge rectification-by-deforming. The former can fill all specified attributes of depth-unavailable pixels for out-of-range too-near/too-far point, diverse noise point, shadow point, reflective point, metal point, and radiant point based on brightness, saturation and depth analysis with no need of actual segmentation and attribute discrimination for targeted depth holes. The latter majorly contain a twisted-edge rectification-by-deforming process for rectifying the twisted depth edges. All of operators in the proposed scheme are well-causally arranged to lead heavy-defect Kinect depth-maps to acquire efficient and rational restoration. Basically, the proposed scheme can tackle most troublesome defects for a Kinect depth map, making the achievement of depth consistency in successive depth maps be not complex. This investigation paves a very low-cost way of promised 3D imaging in use of a sole Kinect.