论文部分内容阅读
针对红外线列和非制冷型焦平面成像系统存在列向条纹非均匀性的现象,提出了一种基于转向核的单帧条纹非均匀性校正新算法。首先,根据图像边缘的梯度特性确定转向核邻域主方向和大小,然后将每列像素的转向核估计值作为该列像素的期望值,以列向条纹非均匀性作为约束进行最小二乘拟合,计算得到每列像元的校正参数,并以一定概率接受最小二乘拟合的结果。当误差函数满足事先设定阈值时,单帧图像的非均匀性校正完成。实验结果表明,该算法具有稳定的收敛性,与同类算法相比能够更有效抑制条纹非均匀性,并且能够保留更多的图像边缘信息。
Aiming at the phenomenon of non-uniform fringe fringes in infrared and uncooled focal plane imaging systems, a new single fringe fringe nonuniformity correction algorithm based on turning nuclei is proposed. Firstly, the main direction and size of the nucleus of the turning nucleus are determined according to the gradient characteristics of the edge of the image, and then the estimated value of the turning nucleus of each column is used as the expected value of the pixel in the column. The least square fitting , Calculate the correction parameters of each pixel, and accept the result of least square fitting with a certain probability. When the error function satisfies the preset threshold, the nonuniformity correction of the single frame image is completed. The experimental results show that the proposed algorithm has a stable convergence and can suppress the fringe non-uniformity more effectively than other similar algorithms, and can preserve more image edge information.