針對高光譜圖像的頻譜簽章之非區域均值演算法

A spectral signature based non-local mean for hyperspectral image denoising

鄧至亨(Chih-Heng TENG)

指導教授 ： 林茂昭

共同指導教授 ： 瑪莉 夏貝(Marie Chabert)

國立臺灣大學/電機資訊學院/電信工程學研究所/碩士(2017年)

https://doi.org/10.6342/NTU201704330

摘要

此論文中介紹一種利用電磁頻譜特徵來除噪的新方法，此法名為高光譜非區域均值去噪演算法。傳統上，為高光譜圖像除噪的方法分成兩大類，一種是利用高光譜上的資訊，另外一種則是利用圖像區域上的資訊，此文中的演算法同時利用頻譜上的資訊與圖像區域上的資訊來除噪。利用圖像區域資訊來除噪的演算法，如平滑濾波器、非區域均值方法、非區域貝氏方法。而利用高光譜資訊來除噪的有主值分析(PCA)、最小噪音成分(MNF)、高光譜圖像最小誤差子空間分析(HySime)。此文中提出之方法企圖結合兩種除噪方法之優點並有以下三點主要貢獻。第一，減少演算法之複雜度。第二，關鍵參數之選取方法。第三，比較並融合既有演算法，並且得到實際測試之結果。

關鍵字

高光譜圖像 ； 除噪 ； 非區域均值 ； 高光譜非區域均值

並列摘要

A new spectral signature method for hyperspectral images denoising named as hyperspectral non-local mean is proposed in this thesis. This method uses spectral information and spatial information to denoise hyperspectral images. Traditionally, spectral information and spatial information are used separately. Thus, there are two different groups of methods to denoise hyperspectral images, spatial algorithms and spectral algorithms. The spatial denoising methods such as smoothing filter, non-local mean and non-local Bayesian consider the correlation in an image. The spectral denoising methods such as PCA (Principal component analysis), HySime (Hyperspectral subspace identification by minimum error) and MNF (Minimum noise fraction) consider the correlation in spectral. Hyperspectral non-local mean takes the advantages of these two groups of algorithms and processes spectral information and spatial information in the same time. Our contributions are 1) reduction of the processing complexity of algorithm. 2) choice of the proper algorithm parameters according to the properties of hyperspectral images. 3) combination and comparison with state-of-the-art.

並列關鍵字

Hyperspectral images ； denoising ； non-local mean ； hyperspectral non-local mean

參考文獻

[3] José M. P. Nascimento, and José M. Bioucas Dias. “Vertex Component Analysis: A Fast Algorithm to Unmix Hyperspectral Data.” IEEE Transactions on Geoscience and Remote Sensing, VOL. 43, NO. 4, pp. 898-910. April 2005.

[5] Guangchun Luo, Guangyi Chen, Ling Tian, Ke Qin, and Shen-En Qian. ‘’Minimum Noise Fraction versus Principal Component Analysis as a Preprocessing Step for Hyperspectral Imagery Denoising.’’ Canadian Journal of Remote Sensing, VOL. 42, NO. 2, pp. 106-116, April 2016.

[6] Lindsay I Smith. “A tutorial on Principal Components Analysis.” February 2002.

[7] José. Bioucas-Dias and José M. P. Nascimento. “Hyperspectral Subspace Identification.” IEEE Transactions on Geoscience and Remote Sensing, VOL. 46, NO. 8. pp. 2435-2445, August 2008.

[8] Zhou Wang, Alan C. Bovik. “A universal image quality index.” IEEE Signal Processing Letters, VOL. 9, NO. 3, pp. 81-84. March 2002.