研究目的
To develop a complete system for acquiring, processing, and rendering panoramic light field stills for VR, addressing challenges in acquisition complexity, data size, rendering quality, and speed.
研究成果
The system successfully provides high-quality light field acquisition, processing, and rendering for VR, with efficient compression and real-time performance. Future work includes extending to light field video and improving hardware and compression methods.
研究不足
The viewing volume is limited, leading to a constraining experience; artifacts remain around depth edges and glossy surfaces; the system is limited to static scenes; offline processing requires extensive cloud resources.
1:Experimental Design and Method Selection:
The system includes two novel light field camera rigs for acquisition, a real-time reconstruction algorithm using per-view geometry and disk-based blending, prefiltering to reduce artifacts, and a modified VP9 codec for compression.
2:Sample Selection and Data Sources:
Light field datasets are acquired using the camera rigs from various scenes (e.g., Living Room, Gamble House).
3:List of Experimental Equipment and Materials:
GoPro Hero4 cameras, Sony a6500 mirrorless cameras, Rokinon fisheye lenses, custom camera rigs with stepper motors and PIC boards.
4:Experimental Procedures and Operational Workflow:
Images are captured, processed through a cloud-based pipeline (calibration, synthesis, subsampling, equiangular conversion, geometry generation, prefiltering, color enhancement, compression), and rendered in real-time using OpenGL on VR hardware.
5:Data Analysis Methods:
Performance is evaluated based on rendering quality (PSNR), file size, and speed (frame rates).
独家科研数据包,助您复现前沿成果,加速创新突破
获取完整内容-
NVIDIA GeForce GTX 1060
GTX 1060
NVIDIA
GPU used in the rendering system for real-time light field reconstruction.
-
GoPro Hero4
Hero4
GoPro
Used in the 16xGoPro rig for capturing light field images.
-
Sony a6500
a6500
Sony
Used in the 2xDSLR rig for high-quality image capture with electronic shuttering.
-
Rokinon fisheye lens
8mm
Rokinon
Attached to Sony a6500 cameras for wide-angle outward pointing.
-
VP9 codec
WebM
Modified for light field compression to provide high quality with real-time random access.
-
Oculus Rift
Oculus
Consumer VR headset used for rendering and displaying light field views.
-
HTC Vive
HTC
Mentioned as a consumer VR headset supporting positional tracking.
-
Intel Core i7-7700HQ
i7-7700HQ
Intel
CPU used in the rendering system for tile decoding and processing.
-
登录查看剩余6件设备及参数对照表
查看全部