摘要
文中把MAP联合求解模型应用到遥感图像序列的超分辨率重建中来,并针对其在边缘保持能力和收敛复杂度方面的局限性,做出如下的改进:运动估计阶段使用SIFT特征算子和光流法相结合提取运动矢量;引入一个二进制掩膜变量,减少模糊边缘效应;迭代重构采用EM算法,使最终结果收敛到一个稳定的最优解。最后的实验结果也表明,针对具有挑战性的低分辨率遥感图像序列的输入,我们的方法可以产生更加清晰和更高分辨率的重建结果。
In this paper,the joint MAP model is applied to the super-resolution reconstruction of the remote sensing image sequence,and the following improvements are made to its limitations in the edge preserving and the convergence complexity:SIFT feature operator and optical flow methods are combined to extract the motion vector in motion estimation stage;introducing a binary mask variable to reduce the fuzzy edge pixels;we use EM iterative algorithm to converge to a stable optimal solution. The final experimental results also show that our approach can produce a clearer and higher resolution reconstruction result for the input of challenging low resolution remote sensing image sequences.
作者
胡龙珍
尹增山
高爽
HU Long-zhen;YIN Zeng-shan;GAO Shuang(Shanghai Institute of Microsystem and Information Technology,Chinese Academy of Sciences,Shanghai 200050,China;Shanghai Engineering Center for Microsatellites,Shanghai 201203,China;University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《电子设计工程》
2018年第21期180-184,189,共6页
Electronic Design Engineering