基于MAP估计的遥感图像频域校正超分辨率算法  

MAP based remote sensing image super-resolution with frequency domain correction

作　　者：王静[1] 章世平[1] 孙权森[1] 夏德深[1] 

机构地区：[1]南京理工大学计算机科学与技术学院,南京210094

出　　处：《东南大学学报（自然科学版）》2010年第1期84-88,共5页Journal of Southeast University：Natural Science Edition

基　　金：香港特区政府研究资助局资助项目(CUHK/4180/01E);国家自然科学基金资助项目(60773172)

摘　　要：为解决超分辨复原中遥感图像调制传递函数过零点引起的病态问题,提出了一种基于最大后验概率估计,利用遥感系统成像模型对图像进行频域校正的遥感图像超分辨率复原算法.该算法首先假设图像满足高斯马尔科夫先验模型,使用最大后验概率估计法实现图像频谱外推.然后根据成像模型对外推的频率分量进行频域校正,去除调制传递函数过零点附近的伪信息.最后使用优化最小化方法完成数值求解,降低了计算复杂度.实验结果证明,该算法考虑了图像成像模型本身特性,对合成图像和各种地物条件下的遥感图像都能取得快速有效的超分辨复原效果.To solve the ill-posed problem imported by zero-crossing of modulation transfer function （MTF） in the super-resolution restoration, a novel MRF-MAP （maximum a posteriori） based super- resolution restoration methods with acquisition system modeling based frequency domain correction is proposed. The Huber function is used in a Gaussian MAP solutions and completes spectral extrapolation. Markov random field（GMRF） to operate stable Then a frequency domain correction algorithm is adopted for the geometrical characteristics of the remote sensing system to revise the frequency spectrum above the cut-off frequency, A majorization-minimization approach is used to reduce the computational complexity. Experimental results indicate that by considering the modeling of image acquisition systems the proposed method can achieve high-quality super-resolution image for both composite image and remote sensing images with different physiognomy.

关 键 词：遥感图像 超分辨率复原 马尔科夫随机场 Huber函数 调制传递函数 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]