%0 Journal Article
%T 基于有界高斯混合模型的高光谱图像去噪方法<br>Hyperspectral Image Denoising Method via Bounded Gaussian Mixture Model
%A 刘昊
%A 贾小宁
%A 成丽波
%A 李喆
%J Advances in Applied Mathematics
%P 8473-8484
%@ 2324-8009
%D 2022
%I Hans Publishing
%R 10.12677/AAM.2022.1112896
%X 针对高光谱图像的噪声去除问题，本文提出了一种基于有界高斯混合模型的高光谱图像去噪方法。在该方法中，我们使用张量表示高光谱图像，并对其进行Tucker分解，最后采用有界高斯混合模型对噪声进行拟合，从而将图像的固有特征和噪声建模相结合。我们将图像和噪声的先验信息表述为一个完整的贝叶斯模型，并设计了变分贝叶斯算法来封闭更新模型中所涉及的变量。最后，我们将本文所提出的去噪算法与其他算法进行对比，验证了该方法的先进性。<br />
To solve the problems of hyperspectral images denoising, we proposed a hyperspectral image de-noising method which is based on bounded Gaussian mixture model in this paper. In this method, we denote hyperspectral images as tensors and use Tucker decomposition to decompose the image tensor into low-rank tensor, at last, we use bounded Gaussian mixture model to capture the noise, in which way we integrate intrinsic image characterizations and noise modeling. Then we describe image and noise priors as a full Bayesian model, and design a variational Bayesian algorithm to in-fer all involved variables by closed-form equations. At last, we compare the denoising method pro-posed in this paper with other algorithms and experiment results prove that our method is state-of-the-art.
%K 高光谱图像，低秩张量分解，贝叶斯模型，变分贝叶斯算法<br>Hyperspectral Images
%K Low-Rank Tensor Decomposition
%K Bayesian Model
%K Variational Bayesian Al-gorithm
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=58783