研究目的
To propose a novel approach for producing integral images using a stereo-plenoptic camera system for full-parallax 3D display, enhancing computation speed with GPU acceleration.
研究成果
The stereo-plenoptic camera system successfully enhances integral image quality and computation speed through GPU acceleration, enabling full-parallax 3D display. Future work should focus on real-time implementation with advanced cameras and improved registration accuracy for non-rigid objects.
研究不足
The system requires calibration to address image distortions; working distance is limited by camera optical properties; real-time implementation is not achieved; full-parallax effect demonstration relies on monocular camera recordings rather than direct binocular observation.
1：Experimental Design and Method Selection:
Utilized a stereo configuration with a commercial plenoptic camera (Lytro Illum) to capture spatial and angular information. Applied depth map estimation from plenoptic images using Jeon's approach, point cloud composition, and registration with Iterative-Closest-Point (ICP) algorithm. Integral image generation via virtual pinhole array (VPA) back-projection with GPU acceleration for parallel computation.
2：Sample Selection and Data Sources:
Captured scenes using the plenoptic camera in different stereo positions; used calibration and rectification processes to handle image distortions.
3：List of Experimental Equipment and Materials:
Lytro Illum plenoptic camera, camera slider, tripod, Samsung SM-T700 tablet as screen, Fresnel Technology MLA (Model 630, f=3.3 mm, p=1.0 mm), NVIDIA GPU (GeForce GTX 870 M), Intel i7 CPU.
4：3 mm, p=0 mm), NVIDIA GPU (GeForce GTX 870 M), Intel i7 CPU.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: Calibrated and rectified plenoptic images, extracted sub-aperture images, estimated depth maps, composed point clouds, registered point clouds with ICP, generated integral images using VPA back-projection with GPU acceleration, and displayed on an integral imaging monitor.
5：Data Analysis Methods:
Compared depth maps and integral image quality; measured computation times for CPU vs. GPU implementations with various interpolation indices.
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
