基于PCA的相机响应函数模型标定算法

许多计算机视觉应用的算法都需要对拍摄场景高动态范围的幅亮度信息进行精确的测量，成像系统的相机响应函数能够建立拍摄图像强度信息与场景辐亮度之间的严格映射关系，是高动态范围图像融合的关键技术。文中分析相机响应曲线的共同特点，结合相机响应函数固有的约束条件，建立相机响应函数的理论空间模型。首先，利用主成分分析法对已有的相机响应数据库进行分析，结合相机响应函数的约束条件建立响应函数的低参数经验模型；然后，根据输入图像选择合适的参数数量；最后，利用不同曝光量的输入图像通过最小二乘法求解建立响应函数模型的系数，从而对相机响应函数进行标定。该算法能够通过对少量的采样点进行插值获得精确的相机响应函数，同时能够对任意的场景通过拍摄多曝光量图像精确地标定相机响应函数。通过对实际拍摄的图像进行相机响应函数标定实验，验证了该算法的有效性，并证明该算法保持高精度的同时计算效率也较高。

Calibration camera response function model algorithm based on principal component analysis

Many computer vision algorithms need to measure the scene radiance accurately, and the camera response function can achieve this result by establishing the mapping between image brightness and scene radiance. Camera response function calibration is the key to high dynamic range image fusion. The properties that all camera response functions share were analyzed in the paper, which helps us to find the constraints that any camera response function must satisfy and establish the theoretical space model of camera response function. Firstly, the database of real-world camera response functions was analyzed by principal component analysis algorithm and low-parameter empirical model of response was established combined with constraints; secondly, appropriate parameter number was chosen according to input images; finally, the coefficients to the low-parameter empirical model of camera response function was solved by least square method. The algorithm proposed in this paper could establish camera response function of the imaging system accurately by interpolating to sparse samples or multiple images with different exposures in arbitrary environment. The effectiveness of this camera response function calibration algorithm was verified by different experiments, which proved high-accuracy and high computational efficiency of this algorithm.