多尺度压缩感知框架下的遥感图像超分辨率重建
发布时间：2018-11-27 20:07
【摘要】：传统的基于压缩感知的超分辨率重建算法将图像看作单尺度,并没有考虑不同尺度的图像块可能包含不同的判别信息.为了有效利用遥感图像的尺度特性,提出了一种多尺度压缩感知框架下的遥感图像超分辨率重建算法.首先通过图像块聚类构建多尺度训练样本集,接着运用Fisher判别准则学习包含遥感图像类别信息的判别字典,然后根据压缩感知中测量矩阵的构造方式估计低分辨率图像的获取过程,最后结合判别字典依次重建多尺度模式下的各子区域图像.实验结果证明了将多尺度压缩感知引入图像超分辨率重建的有效性,提出的算法在视觉效果和评价指标上均优于现有的几种算法.
[Abstract]:The traditional super-resolution reconstruction algorithm based on compression perception regards the image as a single scale and does not consider that the image blocks with different scales may contain different discriminant information. In order to effectively utilize the scale characteristics of remote sensing images, a super-resolution reconstruction algorithm for remote sensing images based on multi-scale compression perception is proposed. Firstly, the multi-scale training sample set is constructed by image block clustering, and then the discriminant dictionary which contains the classification information of remote sensing image is learned by using Fisher discriminant criterion. Then the acquisition process of low-resolution image is estimated according to the construction of measurement matrix in compression perception, and the sub-region images in multi-scale mode are reconstructed in turn by combining the discriminant dictionary. The experimental results demonstrate the effectiveness of introducing multi-scale compression perception into image super-resolution reconstruction. The proposed algorithm is superior to the existing algorithms in terms of visual effect and evaluation index.
【作者单位】： 南京理工大学计算机科学与工程学院;
【基金】：国家自然科学基金项目(61273251) 民用航天技术“十二五”预先研究项目(D040201)
【分类号】：TP751
本文编号：2361901
[Abstract]:The traditional super-resolution reconstruction algorithm based on compression perception regards the image as a single scale and does not consider that the image blocks with different scales may contain different discriminant information. In order to effectively utilize the scale characteristics of remote sensing images, a super-resolution reconstruction algorithm for remote sensing images based on multi-scale compression perception is proposed. Firstly, the multi-scale training sample set is constructed by image block clustering, and then the discriminant dictionary which contains the classification information of remote sensing image is learned by using Fisher discriminant criterion. Then the acquisition process of low-resolution image is estimated according to the construction of measurement matrix in compression perception, and the sub-region images in multi-scale mode are reconstructed in turn by combining the discriminant dictionary. The experimental results demonstrate the effectiveness of introducing multi-scale compression perception into image super-resolution reconstruction. The proposed algorithm is superior to the existing algorithms in terms of visual effect and evaluation index.
【作者单位】： 南京理工大学计算机科学与工程学院;
【基金】：国家自然科学基金项目(61273251) 民用航天技术“十二五”预先研究项目(D040201)
【分类号】：TP751
【相似文献】
相关期刊论文 前2条
1 邹珍军;孙彦荣;;基于四叉树理论的人造地物遥感图像超分辨率技术方法[J];信息与电脑(理论版);2010年04期
2 ;[J];;年期
相关硕士学位论文 前1条
1 王欢;目标识别的遥感图像超分辨率方法研究[D];哈尔滨师范大学;2014年
，本文编号：2361901
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2361901.html
