基于改进NAS-RIF算法的遥感噪声图像自适应复原  

Self- adaptive restoration for remote sensing noise images based on improved NAS- RIF algorithm

作　　者：张凡[1] 

机构地区：[1]河南经贸职业学院信息管理系,郑州450018

出　　处：《国土资源遥感》2015年第2期105-111,共7页Remote Sensing for Land & Resources

基　　金：2013年度河南省高等学校教学工程项目"2013年度河南省高等学校教学团队-计算机应用"(编号:教高[2013]589号)资助

摘　　要：实现对遥感噪声图像的有效复原是遥感图像处理的一项重要研究内容。在对非负支撑域有限递归逆滤波(non-negativity and support constraints recursive inverse filtering,NAS-RIF)算法深入研究的基础上,提出一种基于改进自适应NAS-RIF算法的遥感噪声图像复原方法。该算法针对经典NAS-RIF算法存在的缺陷,首先对含有椒盐噪声和高斯白噪声的遥感图像采用自适应伪中值滤波算法进行预处理,以尽可能排除图像中噪声的干扰;然后结合图像的灰度值,从算法支撑域和背景灰度值2个方面加以改进;最后对代价函数引入基于目标信息的修正项,改进了经典NAS-RIF算法的代价函数;与对数函数复合,使得改进后NAS-RIF算法的代价函数具有良好的收敛性;并采用共轭梯度法对改进自适应NAS-RIF算法进行整体优化。对仿真实验结果进行的主观和客观分析表明,本文算法的性能优于经典NAS-RIF算法、已有的改进NAS-RIF算法以及小波阈值去噪方法,能够胜任遥感噪声图像的复原处理。Image restoration is an important research content in remote sensing noise image processing. In order to deal with the remote sensing image effectively, this paper proposes a new improved self -adaptive NAS -RIF algorithm based on the research on the basic principle of the non -negativity and support constraints recursive inverse filtering （NAS-RIF）. With the purpose of tackling the defects of the original NAS-RIF algorithm,the author first filtered the image with pepper and salt noise as well as white Gaussian noise by the self -adaptive pseudo-median filtering algorithm so as to eliminate the noise in the image as much as possible, then improved the original NAS - RIF algorithm effectively from two aspects of support domain and background gray value, in combinatiuon with the gray values of the image and finally introduced the correction term based on the target information to the cost function so as to improve the classic cost function of the original NAS -RIF algorithm. Aimed at improving the convergence of the cost function of the improved NAS-RIF algorithm,the author combined the logarithmic function and adopted the conjugate gradient method to optimize the improved NAS-RIF algorithm. Subjective and objective analysis of the simulation experimental results shows that the performance of the improved NAS-RIF algorithm proposed in this paper is better than that of the original NAS -RIF algorithm and some available improved NAS -RIF algorithms as well as the wavelet threshold denoising method, suggesting that this means is suitable for the restoration process of the remote sensing noise image.

关 键 词：图像复原 自适应伪中值滤波 非负支撑域有限递归逆滤波(NAS-RIF) 改进的自适应NAS-RIF算法 代价函数 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置] TP391[自动化与计算机技术—控制科学与工程]