中文摘要：

针对由传统的相机响应函数标定法合成的高动态范围图像质量较差的问题,提出一种有效的高动态范围成像算法.首先根据多曝光图像灰度变化的特点,通过三次样条插值将相机响应函数标定问题转化为求解一个线性三对角线性方程组;然后根据拍摄场景动态范围的变化情况,通过手动调整曝光的方式使多曝光图像大部分像素的正序或倒序的灰度变化趋于恒定,再通过逐点递推的方式求出三次样条函数各离散端点的二阶导数,拟合出相机响应曲线;最后根据已选定的基准点并结合已标定的相机响应曲线恢复出单位曝光度下像素对应的曝光时间,即真实的亮度辐射值.实验结果表明,三次样条插值能够有效地提高图像的局部细节以及图像整体的清晰度,递推法简化了求解线性方程组烦琐的计算步骤,降低了算法的整体运算时间.

英文摘要：

Due to the limited quality of high dynamic range image which is restored by traditional calibration methods, this paper proposes a novel high dynamic imaging method to improve the quality of image. By cu- bic spline interpolation method, our algorithm converts the calibration of camera response function into solving a system of linear equations primarily. According to the variation of dynamic range, our algorithm makes an adjustment to the ratio of exposure manually in order to guarantee that the variation of the gray value is tending towards stability mostly. Afterwards, a recursive method is proposed to calculate the sec- ond-order derivative of each end point of spline function which is the essential condition of calibration. At last, the radiance of each pixel is recovered according to the reference point which is chosen previously and the result of calibration. Experiment results show that cubic spline interpolation is able to improve the local detail and global clarity. The proposed recursive method reduces the complexity of solving linear system of equations, so that real-time performance can be achieved.