基于Gini指数的高光谱图像空、谱相关稀疏性分析及压缩感知联合重建方法研究

项目名称： 基于Gini指数的高光谱图像空、谱相关稀疏性分析及压缩感知联合重建方法研究

项目编号： No.61301217

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 无线电电子学、电信技术

项目作者： 李星秀

作者单位： 南京理工大学

项目金额： 24万元

中文摘要： 传统的高光谱成像技术依据Nyquist采样定理进行采样，获取高空间分辨率的高光谱图像需要大数据量的采样和传输，这给低功耗和带宽有限的实用高光谱遥感平台硬件系统带来巨大的压力。本项目以高光谱图像为研究对象，以压缩感知高光谱图像联合重建为科学问题，通过深入分析高光谱图像的空间维和光谱维相关性，研究高光谱图像的空、谱维稀疏表示系数联合聚类结构特征及其基于Gini指数的度量方法，建立基于空、谱维相关稀疏性先验的压缩感知高光谱图像联合重建模型和算法，旨在从根本上进一步减少高光谱图像的采样数据，降低高光谱图像的获取成本。本课题对丰富拓展稀疏表示、压缩感知理论和算法具有十分重要的理论意义，同时将为设计新型的低功耗CS高光谱成像系统奠定良好的理论基础，具有十分广阔的应用前景。

中文关键词： 高光谱图像；低秩；TV正则化；空谱稀疏性；压缩感知

英文摘要： The conventional hyperspectral imaging technique is based on Nyquist sampling theorem. So acquiring the high-spatial-resolution hyperspectral images needs to sample and transmit a large amount of data, which brings serious challenges to the pratical hyperspectral remote sensing platform processing systems where the power consumption and bandwidth are tight constraints. In this subject, we consider the compressed sensing hyperspectral images joint reconstruction problem for reducing the hyperspectral images sampling data essentially. In particular, by analyzing the spatial and spectral correlation of hyperspectral images deeply, a gini index based measure of spatial and spectral correlation sparsity prior is studied. Finally, a set of novel compressed sensing reconstruction model and algorithm for hyperspectral images are established. This project will enrich and extend the sparse representation and the compressed sensing theory and algorithm in the theoretical sense, and promote the design and practical application of new compressed sensing hyperspectral imaging system.

英文关键词： Hyperspectral images；low-rank；total variation regularization；spatial-spectral sparsity；compressed sensing