论文部分内容阅读
提出了一种基于场景的空间投影估计和时域迭代百均匀性校正算法(SPETI).这种方法主要是通过投影估计来估算全局位移,并且在连续多帧图像中进行迭代计算.首先,为配准准则设计了一个新的投影估计;然后,计算相邻帧并进行迭代计算,以此来获得较快的固定图案噪声(FPN)收敛,同时减少鬼影的产生;最后,将本算法移植至基于FPGA的硬件系统中.在连续单调运动的系统中测试算法的各项指标,并且将实际的校正效果予以展示.为了与自适应最小均方差算法和全变差算法进行比较,对一个完全干净的红外图像序列添加了人为的非均匀性.将固定标准增益和偏置非均匀性添加到图像序列上,以测试迭代次数与非均匀性等级的关系.
A scene-based spatial projection estimation and time-domain iterative uniformity uniformity correction algorithm (SPETI) is proposed, which estimates the global displacement by projection estimation and iteratively computes in continuous multi-frame images.First, The registration criterion designs a new projection estimation. Then, the neighboring frames are calculated and iteratively calculated, so as to obtain fast fixed pattern noise (FPN) convergence and reduce the generation of ghosting. Finally, the algorithm is ported To FPGA-based hardware system.In the continuous monotone motion system, we test all the indexes of the algorithm and display the actual correction effect.In order to compare with the adaptive minimum mean square deviation algorithm and total variation algorithm, Artificial inhomogeneities were added to the clean infrared image sequence Fixed fixed and offset inhomogeneities were added to the image sequence to test the relationship between the number of iterations and the non-uniformity level.