基于非下采样Shearlet变换与双重制约的遥感图像融合算法  

Remote sensing image fusion algorithm based on nonsubsampled shearlet transform coupling double restriction model

作　　者：史丽燕 王合闯 SHI Liyan;WANG Hechuang(College of Information Engineering,Zhengzhou Vocational University of Information and Technology,Zhengzhou 450046,China;College of Information Engineering,North China University of Water Resources and Electric Power,Zhengzhou 450046,China)

机构地区：[1]郑州信息科技职业学院信息工程学院,河南郑州450046 [2]华北水利水电大学信息工程学院,河南郑州450046

出　　处：《光学技术》2021年第3期352-358,共7页Optical Technique

基　　金：国家自然科学基金(51609086);河南省科技攻关计划项目(182400410104,142102210041);河南省软科学研究计划项目(202400410333)。

摘　　要：为了提高遥感图像的融合质量,使其兼顾较好的纹理特征与对比度,提出了非下采样变换耦合双重制约模型的遥感图像融合算法(NSST)。通过IHS模型来解析多光谱图像(MS),提取其对应的强度(I)、色调(H)、饱和度(S)成分;借助NSST,从I成分和全色图像(PAN)中解析出低频和高频系数;通过信息熵和均值模型,计算出图像富含的信息及亮度丰富度,以完成低频系数的融合。利用高频系数与方向矩阵的卷积运算,得出图像的纹理特征,计算图像的标准差,获取对比度信息。在联合纹理特征和对比度,构造双重制约模型,完成高频系数的融合。再对融合系数完成逆NSST和逆IHS运算,得出融合图像。实验数据表明,较现有的融合技术而言,所提算法的融合图像含有更为丰富的纹理与更高的对比度。In order to make the fused remote sensing image have excellent contrast while highlighting the texture features, a double restriction model to fuse the remote sensing image in the non subsampling shearlet transform domain is designed. Firstly, the multispectral image is analyzed by IHS model, and divided into intensity, hue and saturation components. Then, with the help of NSST, the low and high frequency coefficients of I component and panchromatic image are analyzed. Finally, through the information entropy and mean value model, the image rich information and brightness richness are calculated, and the low-frequency fusion coefficient is calculated. Using the convolution operation of image and direction matrix,the texture features of image are calculated.By calculating the standard deviation of image,the contrast information of image is obtained.Using the texture and contrast information of image,the double restriction model is constructed,and the highfrequency fusion coefficient is calculated with it,and then the fusion image is obtained.The experimental results show that the fusion performance of this algorithm is better than the existing algorithm,and the fusion image texture is more prominent,the contrast is better,and the information richness is higher.

关 键 词：纹理特征 遥感图像融合 全色图像 多光谱图像 高频系数 低频系数 卷积运算 双重制约 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]