基于混合稀疏基字典学习的微波辐射图像重构方法

发布时间：2018-04-10 11:24

本文选题：微波辐射图像 + 超稀疏干涉测量　；参考：《电子与信息学报》2016年11期

【摘要】：目前的微波辐射测量成像系统在一次观测中所采集的数据量大,基于奈奎斯特空间采样及常规微波辐射图像重构方法难以实现高分辨率要求。该文针对微波辐射干涉测量在频域中进行,采用傅里叶最优随机抽取的超稀疏干涉测量(低于奈奎斯特采样)对微波辐射图像进行线性压缩投影,降低数据采样。考虑微波辐射图像在总体差分域和小波中都具有可压缩特性,提出总体差分和小波混合正交基的K-SVD字典学习微波辐射图像重构模型,利用Bregman和交替迭代算法求解该模型,重构线性压缩投影信息从而获得微波辐射图像。仿真实验表明,该文提出的算法在微波辐射图像重构效果、噪声稳定性上优于DLMRI算法和Grad DLRec算法。
[Abstract]:Microwave radiometric imaging system is collected in a large amount of data in the observation, Nyquist space sampling and conventional microwave radiation image reconstruction method is difficult to achieve high resolution. In this paper, based on the requirements of microwave radiation in the frequency domain interferometry, ultra sparse interferometric Fu Li optimal random leaf (below the Nyquist sampling) image compression of linear projection of the microwave radiation, reduce the sampling data. Considering the difference in the overall microwave radiation image domain and wavelet has the compression property, puts forward the overall difference and wavelet mixed K-SVD Dictionary of microwave radiation image reconstruction model is to base, using Bregman and alternating iterative algorithm to solve the model, the reconstruction of linear compression in order to obtain the projection information of microwave radiation image. Simulation experiments show that the proposed algorithm for image reconstruction in the microwave radiation effect, noise The stability is superior to the DLMRI algorithm and the Grad DLRec algorithm.

【作者单位】：华东交通大学信息工程学院;
【基金】：国家自然科学基金(31101081;61162015) 江西省自然科学基金(20161BAB202061)~~
【分类号】：TP751

【相似文献】