真实的高光谱图像(HSI)容易遭受高强度混合噪声的破坏,如何精确地对噪声进行建模在后续处理任务中至关重要。基于非对称拉普拉斯噪声建模方法取得了较好的混合噪声去除效果,该类方法考虑到噪声的重尾性和非对称性,对不同波段的不同噪声进行建模。但忽略了HSI梯度基空间U的内在分布特征,导致噪声残留。针对此问题,提出一种基空间非对称拉普拉斯全变分(BSALTV)的HSI去噪模型。梯度基空间U充分保留了原始梯度图的先验信息,能够更好地反映HSI梯度的稀疏先验分布特征,并且在不同波段上呈现出独特的非对称分布。通过对梯度基U和噪声的非对称分布进行探索,精确挖掘了图像的全局低秩信息和不同波段的噪声分布特点,从而在保持图像边缘和纹理的同时减少噪声,避免了图像失真和过度平滑。最后,通过ADMM算法求解模型,在合成和真实数据集上的大量实验结果表明,所提方法优于对比的其他先进的降噪方法。

Real hyperspectral images (HSI) are vulnerable to high intensity mixed noise, and how to accurately model the noise is very important in the subsequent processing tasks. Based on the asymmetric Laplacian noise modeling method, a good mixed noise removal effect is obtained. This method takes into account the heavy tail and asymmetry of noise and models different noises in different bands. However, the internal distribution characteristics of HSI gradient base space U are ignored, resulting in residual noise. To solve this problem, a HSI denoising model based on asymmetric Laplacian total variation (BSALTV) is proposed. The gradient base space U fully retains the prior information of the original gradient map, and can better reflect the sparse prior distribution characteristics of HSI gradients, and presents a unique asymmetric distribution on different bands. By exploring the asymmetric distribution of gradient base U and noise, the global low-rank information and noise distribution characteristics of different bands of the image are precisely mined, thus reducing noise while maintaining image edges and textures, avoiding image distortion and excessive smoothing. Finally, the model is solved by ADMM algorithm, and a large number of experimental results on synthetic and real data sets show that the proposed method is superior to other advanced noise reduction methods compared.