分块噪声自适应高光谱图像去噪算法研究

高光谱图像在获取过程中容易产生噪音,从而影响了地物空间信息的识别。噪声去除是高光谱图像处理十分必要的步骤。结合低秩矩阵分解理论,在传统奇异值阈值方法的基础上提出基于分块的噪声自适应遥感去噪方法。实验结果证明,该方法运算速度快,并能够有效去除缺失值造成的死线噪声以及高斯噪声,在平均峰值信噪比(MPSNR)和平均结构相似性(MSSIM)上优于Godec算法。

RESEARCH ON THE ALGORITHM OF NOISE-ADAPTIVE HYPERSPECTRAL IMAGE DENOISING BASED ON BLOCK

Hyperspectral images carry abundant spectral information and the spatial information which can be used to identify the objects on the land, so they have a great use in many areas. However, the hyperspectral images are often affected by the noise in the obtaining process, which is not conducive to the subsequent application. Therefore, noise removal is a very necessary step in the processing of hyperspectral image. The clean hyperspectral image data matrix is low rank. Combining with low rank matrix decomposition theory, based on the traditional singular value threshold method, a method of adaptive noise reduction based on block is proposed. Experimental results show that the computing speed of this method is fast, and can effectively remove missing values caused by the deadline noise and Gaussian noise, the results which are compared with Godec algorithm in average peak signal to noise ratio (MPSNR) and mean structure similarity (MSSIM) are better.