View Synthesis: LiDAR Camera versus Depth Estimation

Depth-Image-Based Rendering (DIBR) can synthesize a virtual view image from set of multiview images and corresponding depth maps. However, this requires an accurate map estimation that incurs high compu- tational cost over several minutes per frame in DERS (MPEG-I’s Depth Estimation Reference Software) even by using high-class computer. LiDAR cameras thus be alternative solution to real-time DIBR ap- plications. We compare the quality low-cost camera, Intel Realsense L515 calibrated configured adequately, with MPEG-I’s View Synthesizer (RVS). In IV-PSNR, camera reaches 32.2dB synthesis 15cm baseline 40.3dB 2cm baseline. Though outperforms 4.2dB, latter provides better quality-performance trade- off. visual inspection demonstrates LiDAR’s views have slightly higher than most tested low-texture scene areas, except for object borders. Overall, we highly recommend advanced methods (like DERS) applications. Neverthe- less, delicate calibration multiple tools further exposed paper.