研究目的
To propose a calibration method for focused light field camera based on virtual image points, establishing mappings from object points to image points and solving geometrical parameters using iterative algorithms.
研究成果
The proposed calibration method based on virtual image points is feasible and accurate, with small reprojection errors (average 1.6092 pixels horizontal and 1.8012 pixels vertical). It provides a foundation for computational imaging techniques in light field cameras.
研究不足
The method neglects lens distortions caused by the main lens and microlens, which could affect accuracy. Future work includes modeling these distortions and optimizing focus distances for different microlens types.
1:Experimental Design and Method Selection:
The methodology involves establishing a forward model for light field camera imaging, using pinhole and thin-lens models, and employing the Levenberg-Marquardt algorithm for nonlinear optimization to solve for intrinsic and extrinsic parameters.
2:Sample Selection and Data Sources:
A checkerboard calibration plate with a grid size of 10 mm x 10 mm and 11x8 corner points is used as the object. Raw light field images are captured at a focus distance of 1500 mm with 15 different poses.
3:List of Experimental Equipment and Materials:
A multi-focus light field camera with a prime lens of 75 mm focal length, detector resolution of 7716x5364 pixels, pixel size of
4:12 μm x 12 μm, and a checkerboard of size 120 mm x 90 mm. Experimental Procedures and Operational Workflow:
Data preprocessing to locate microlens centers using white images and Gaussian filtering. Calibration experiments involve capturing images of the checkerboard in various poses, detecting corner points, grouping them, calculating virtual image points, and solving parameter matrices iteratively.
5:Data Analysis Methods:
Reprojection error analysis is performed to validate accuracy, using differences between reprojected and actual virtual corner points coordinates.
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