Snapshot spectral compressive imaging reconstruction using convolution and contextual Transformer  

作　　者：LISHUN WANG ZONGLIANG WU YONG ZHONG XIN YUAN 

机构地区：[1]Chengdu Institute of Computer Application,Chinese Academy of Sciences,Chengdu 610041,China [2]University of Chinese Academy of Sciences,Beijing 100049,China [3]Research Center for Industries of the Future and School of Engineering,Westlake University,Hangzhou 310030,China

出　　处：《Photonics Research》2022年第8期1848-1858,共11页光子学研究（英文版）

基　　金：New Generation of Artificial Intelligence Integration and Application Demonstration of the Chinese Academy of Sciences(RTLZ2021009);Westlake Foundation(2021B1501-2)。

摘　　要：Spectral compressive imaging(SCI) is able to encode a high-dimensional hyperspectral image into a twodimensional snapshot measurement, and then use algorithms to reconstruct the spatio-spectral data-cube. At present, the main bottleneck of SCI is the reconstruction algorithm, and state-of-the-art(SOTA) reconstruction methods generally face problems of long reconstruction times and/or poor detail recovery. In this paper, we propose a hybrid network module, namely, a convolution and contextual Transformer(CCo T) block, that can simultaneously acquire the inductive bias ability of convolution and the powerful modeling ability of Transformer, which is conducive to improving the quality of reconstruction to restore fine details. We integrate the proposed CCo T block into a physics-driven deep unfolding framework based on the generalized alternating projection(GAP) algorithm, and further propose the GAP-CCo T network. Finally, we apply the GAP-CCo T algorithm to SCI reconstruction. Through experiments on a large amount of synthetic data and real data,our proposed model achieves higher reconstruction quality(>2 d B in peak signal-to-noise ratio on simulated benchmark datasets) and a shorter running time than existing SOTA algorithms by a large margin. The code and models are publicly available at https://github.com/ucaswangls/GAP-CCoT.

关 键 词：CONVOLUTION SPECTRAL DETAILS 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]