常规的去噪方法在去除遥感卫星影像噪声时,往往会造成影像模糊和空间分辨率下降。本文将稀疏表示理论应用于遥感卫星影像去噪,提出了一种改进算法,能够保留低频信息不变,仅对影像的高频信息进行稀疏重建。算法核心是按照是否能够从过完备字典中选择较少原子进行稀疏表示的原则来区分高频信息中的有效信息和噪声。通过3组实验对改进算法与传统去噪方法进行对比检测,实验结果表明,改进算法在去除噪声的同时,能较好地突出影像的边缘和细节信息。

Considerable noise is present in some multispectral images acquired by remote-sensing satellites. The current traditional de-noising methods not only fail to completely remove the noise, but also cause image blurring and spatial-resolution degradation. This study aims to mitigate the tradeoff between the removal of noise and the reservation of information.
To solve this problem, we propose an improved and high-performing sparse representation approach that processes the high-frequency portions in the difference images based on the initial image and the Gaussian-filtered image to remove the noise. In this study, sparse representation is applied to the information in a remote-sensing image to accurately represent important information, which includes edge and texture. By contrast, the noise that is mainly concentrated in the high-frequency portion cannot be represented. We used data sparsity to reconstruct the high-frequency portion without noise.
The algorithm completely preserves the low-frequency information and reconstructs the high-frequency information by sparse representation based on whether or not such information can be represented by fewer atoms from the over-complete dictionary.
Theoretical analysis and experimental results show that the proposed method outperforms the traditional de-noising methods and the sparse representation method. In terms of visual quality, the proposed method reconstructs the image with clear color and apparent structure. The results of the objective assessment show that the proposed method can achieve a higher peak signal-to-noise ratio than the other methods and provide a feasible solution to remove noise effectively and considerably highlight the details of the original images.