基于神经网络的红外焦平面光学非均匀性校正改进算法

基于场景的非均匀校正依然是红外领域的一个研究热门。神经网络算法是一种较为典型的场景校正算法。本文主要针对神经网络算法本身不能校正光学引入的非均匀性问题,提出了新的改进算法,通过对神经网络输入层的预处理,消除图像的低频噪声,此外,为了消除预处理对图像对比度的影响,本文增加了神经网络的层数,使用双层神经网络对算法进行更新,从而消除了图像对比度下降的现象。实验结果表明,改进的神经网络算法能够有效的改善图像质量,消除图像中光学引入的非均匀性。

An Improved Algorithm for IRFPA Optical Nonuniformity Correction Based on Neural Networks

Scene-based non-uniformity correction is still a hot topic in the infrared field.A neural network algorithm is a classical scene-based non-uniformity correction algorithm.This article mainly introduced problems where the classical algorithm cannot correct an optical non-uniformity.We propose an improved algorithm based on the preprocessing layer to correct to the low-frequency noise.In order to eliminate the influence of the image contrast,we add a learning layer that can eliminate the image contrast drop phenomenon.The results of the experiment show that the new algorithm can effectively improve the image quality and eliminate non-uniformity introduced by the optics.