基于张量的高维多通道图像压缩感知重建理论与算法及其应用

项目名称： 基于张量的高维多通道图像压缩感知重建理论与算法及其应用

项目编号： No.91538108

项目类型： 重大研究计划

立项/批准年度： 2016

项目学科： 无线电电子学、电信技术

项目作者： 韦志辉

作者单位： 南京理工大学

项目金额： 67万元

中文摘要： 随着新型传感器、计算机通信与网络技术的飞速发展和广泛应用，人们所面临的数据已经不再是简单信号或者图像，而是规模更大、维数更高、结构更复杂的高维多通道数据。本项目通过对基于张量的高维多通道图像的多特征联合几何正则化先验度量与建模，以及张量模式下高维多通道图像的结构化低秩稀疏先验度量与建模，提出高维多通道图像的压缩感知超分辨重建理论和方法，并针对高光谱图像的特点，利用所提出的基于张量的高维多通道图像稀疏重建理论和方法，提出高光谱图像质量提升复原和压缩感知超分辨重建新方法。本项目研究不仅推动压缩感知理论和图像超分辨重建理论在高维多通道图像处理中的发展，也推动张量理论和方法在图像处理领域中的发展，具有十分重要的理论意义；同时，所提出的图像重建模型和算法将不仅可以用于基于高光谱（多光谱）的高分辨率遥感图像重建，还可以在视频监控、多模态生物信息处理等许多重要应用领域发挥重要作用，具有广泛的应用前景。

中文关键词： 高维多通道图像；压缩感知重建；张量分析；几何正则化；地秩稀疏正则化

英文摘要： With the rapid development and wide application of the new type of sensors, computer communication and network technology, the data people have to face is no longer a simple signal or image, but the high-dimensional multi-channel data with larger scale, higher dimension and more complex structure. In this project, we propose a tensor based multi-characteristics joint geometry regularization prior and structured low rank sparse prior modeling method for high-dimensional multi-channel image, and then propose new compressive sensing super-resolution reconstruction theory and algorithms for high-dimensional multimodal image. According to the characteristics of hyperspectral image, by using the proposed tensor based sparse reconstruction theory and method for high-dimensional multi-channel image, we propose new theory and algorithms for the quality- promoting recovery and compression sensing super-resolution reconstruction of hyperspectral image. The research of project can not only promote the development of compression sensing theory and super-resolution reconstruction theory for the high-dimensional multi-channel image, but also promote the development of tensor theory and method in the field of image processing, and therefore it has important theoretical significance; At the same time, the theory and method proposed in this project can be used for the compressive sensing super-reconstruction of the hyperspectral images, also can be used in video monitoring, multimodal biological information processing and so on, and therefore it plays an important role in many important applications, and has wide application prospect.

英文关键词： high-dimensional multi-channel image ;compressive sensing reconstruction;tensor analysis;geometry regularization;low-rank sparse regularization