一种多尺度压缩感知的遥感图像超分辨重建方法  

A multi-scale copressive sensing based super-resolution reconstruction method for remote sensing images

作　　者：杨学峰[1] 赵冬娥[1] YANG Xuefeng;ZHAO Dong’e(School of Information and Communication Engineering,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学信息与通信工程学院,山西太原030051

出　　处：《应用科技》2020年第4期20-25,共6页Applied Science and Technology

基　　金：十三五装备预研基金项目(61404150304).

摘　　要：针对传统基于压缩感知的单字典超分辨方法难以充分描述复杂的遥感图像纹理的问题,提出了一种多尺度残余字典超分辨重建方法。首先对插值图像的高频子带执行Contourlet变换获得多个子频带;然后在各子频带上建立对应子残余字典,并进行字典学习和超分辨重建;最后对高频和低频部分进行融合得到完整的超分辨图像。实验结果表明:与其他相关方法相比,本文方法的超分辨效果无论主观视觉还是客观评价指标都有很大提高。其中客观评价指标,本文方法的峰值信噪比(peak signal-to-noise ratio,PSNR)和结构相似度(structural similarity,SSIM)平均值分别提高6 dB和0.05。对于纹理复杂的遥感图像的超分辨重建场合,更好地满足重建效果和时效的要求,具有重要的理论和应用意义。The traditional compressive sensing based single dictionary super-resolution method can not fully describe the complex remote sensing image texture.To resolve this problem,this paper proposes a multiple residual dictionary method.Firstly,the method extracts the high-frequency part from interpolated image to perform Contourlet transform to obtain multiple directional sub-bands;then establishes corresponding residual sub-dictionary in each sub-band,and perform dictionary learning and super-resolution reconstruction;finally,combines the high-frequency part and lowfrequency part to obtain the whole super-resolution image.The experimental result shows that comparing with other state-of-art methods,the super resolution effect of the proposed method is improved greatly both in subjective evaluation and objective evaluation indexes.For the objective evaluation indexes,the average values of peak signal-to-noise ratio(PSNR)and structural similarity(SSIM)of the proposed method have improved 6 dB and 0.05 respectively.For the super-resolution reconstruction of a remote sensing image with complex texture,it better meets the requirements of good reconstruction effect and fast speed and has important theoretical and practical significance.

关 键 词：遥感图像 超分辨 压缩感知 多尺度变换 残余字典 CONTOURLET变换 稀疏编码 字典学习 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]