基于稀疏正则化结合NLS的超分辨率图像重建

为了保持高光谱(HS)超分辨率重建过程中的频谱一致性和边缘锐度,提出一种基于空间谱结合非局部相似性的超分辨率重建算法。首先,使用HS图像生成模型,采用稀疏正则化解决全色(PAN)图像和HS图像重建的病态问题求逆；然后分析了从高空间分辨率到低空间分辨率数据生成的丰度系数映射；最后利用非局部相似性,设计空间谱联合正则化项。实验结果表明,本文算法重建图像在PSNR,SSIM和FSIM方面明显高于其他优秀算法,在SAM和ERGAS方面明显低于其他优秀算法,在光谱失真方面丢失最少,仅有2%-3%,低于其他算法30%左右,且重建效果更加清晰自然。

Super-Resolution Image Reconstruction Based on Fusion of Sparse Regularization and NLS

To maintain spectral consistency and edge sharpness during the processing of super-resolution reconstruction of hyperspectral (HS) images. A joint super-resolution algorithm based on fusion of space spectrum and non-local similarity (NLS) is proposed. Firstly, HS images are used to generate model, and sparse regularization is used to solve the inversion of the ill problem of the reconstruction of panchromatic (PAN) images and HS images. Then, the generated map of abundance coefficients between spatial high resolution and low resolution is analyzed. Finally, space spectrum joint regularization term is designed by non-local similarity. The proposed method is tested with Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) and Hyperion images. Experimental results show that the reconstructed image by this paper is obviously higher than other good algorithms on PSNR, SSIM and FSIM, and lower than other outstanding algorithms significantly on SAM and ERGAS. Proposed algorithm misses the least spectral with only 2% to 3%, which is 30% lower than other algorithms, and the reconstruction results are more natural and clear.