Liu Zhendong, Guan Hongliang, Ni Qingyang
College of Resource Environment and Tourism, Capital Normal University, Beijing 100048, China.
Engineering Research Center of Spatial Information Technology, Beijing 100048, China.
Sensors (Basel). 2025 Aug 10;25(16):4940. doi: 10.3390/s25164940.
Accurate and robust calibration of multifocal plenoptic cameras is essential for high-precision 3D light field reconstruction. In this work, we propose a blur feature-guided cascaded calibration for the plenoptic camera. First, white images at different aperture values are used to estimate the high-confidence center point and radius of micro-images, and the defocus theory is used to estimate the initial values of the intrinsic parameters. Second, the gradient value is introduced to quantify the degree of blurring of the corner points, which are then divided into three types: clear, semi-clear, and blurred. Furthermore, a joint geometric constraint model of epipolar lines and virtual depth is constructed, and the coordinates of the semi-clear and blurred corner points are optimized in a step-by-step manner by using the clear corner point coordinates. The micro-image center ray projection equation is then devised to assist in the optimization of the microlens array core parameters and establish blur-adaptive credibility weights, thereby constructing a global nonlinear optimization. Finally, the proposed method is tested on both simulated and captured datasets, and the results exhibit superior performance when compared with the established methods described by Labussière, Nousias, and Liu. The proposed method excels in corner feature extraction, calibration accuracy of both internal and external parameters, and calibration sensitivity when applied to multifocal-length light field cameras, highlighting its advantages and robustness.
对于高精度三维光场重建而言，多焦点全光相机的精确且稳健的校准至关重要。在这项工作中，我们提出了一种用于全光相机的模糊特征引导级联校准方法。首先，使用不同光圈值下的白色图像来估计微图像的高置信度中心点和半径，并利用散焦理论来估计内参的初始值。其次，引入梯度值来量化角点的模糊程度，然后将其分为清晰、半清晰和模糊三种类型。此外，构建了极线和虚拟深度的联合几何约束模型，并利用清晰角点坐标逐步优化半清晰和模糊角点的坐标。接着设计微图像中心光线投影方程，以辅助微透镜阵列核心参数的优化并建立模糊自适应可信度权重，从而构建全局非线性优化。最后，在模拟数据集和采集数据集上对所提出的方法进行了测试，与Labussière、Nousias和Liu所描述的现有方法相比，结果显示出卓越的性能。所提出的方法在角点特征提取、内外参数校准精度以及应用于多焦距光场相机时的校准灵敏度方面表现出色，突出了其优势和稳健性。
