基于全维度独立采样机制的高光谱快速成像及高维去噪方法研究

Efficient application of both spatial and spectral correlation for improving the resolution of imaging and simultaneously easing the difficulty of big data sampling, transmission, storage and processing, is a key point of hyperspectral researches. Compressive sensing (CS) theory can utilize the compressibility of data to implement the sampling and compression at the same time, which is beneficial to reducing the pressure of sampling hardware and transmission bandwidth. This project propose to develop a new scheme for hyperspectral compressed imaging based on the so called independent sampling in whole dimensions to achieve fast imaging. The fast high-dimensional reconstruction algorithms based on the independent sampling in whole dimensions serves as the core of this project, which is in conjunction with the high-dimensional de-noising algorithms to improve the accuracy of imaging. The new scheme is composed of compressive sampling, data reconstruction and high-dimensional de-noising in hyperspectral imaging. Independent sampling scheme in whole dimensions with application to hyperspectral fast imaging will bring us the advantages as high efficiency of sampling rate, fast and flexible reconstruction, high imaging resolution and high accuracy.

如何充分利用高光谱数据空间维和光谱维的相关性提高成像分辨率，同时缓解数据量大而造成的采样、传输、存储和处理困难问题，是高光谱成像研究中亟待解决的问题。压缩感知理论可以利用数据的可压缩性，将采样与压缩同时进行，在一定程度上缓解了采样硬件和传输带宽的压力。本项目以高光谱成像为研究背景，以基于全维度独立采样机制的高光谱快速压缩成像为研究目标，重点研究高维数据重构算法，结合高光谱图像的高维去噪方法，形成压缩采样、数据重构、高维去噪的系统方法，实现快速精确的高光谱压缩感知成像。构建全维度独立采样机制并将其应用于高光谱快速成像，具有压缩效率高、数据重构快速灵活、成像分辨率高等具有现实意义的优势。

本项目重点研究了基于全维度独立采样机制的高光谱快速成像方法及高维去噪方法。构建了全维度独立采样机制，对高光谱图像的全部维度同时进行压缩采样，解决了高光谱大数据传输、存储和处理对采样硬件和带宽造成巨大压力问题。提出了基于截断式张量贝叶斯的高光谱图像重建方法，在全维度独立采样机制的条件下，实现快速高精度的图像重建，为高分辨率高光谱成像提供研究基础。提出了多向低秩建模和空间-光谱全变分模型，在去除高光谱图像混合噪声方面的优越性能，为后续图像处理、分析及应用提供有利的条件。本项目研究为高光谱后续应用提供了良好的前提条件，并为探索具有压缩采样功能的新型高光谱探测载荷物理模型提供了坚实的研究基础。