The Research Of Hyperspectral Image Denoising Based On Superpixel Segmentation And Low-Rank Representation

Hyperspectral image(HSI) has various characteristics such as numerous bands, large amounts of data, high spectral resolution and so on, therefore, HSI with a wide range of applications in features detection, environmental monitoring, target identification and so on. However, since the acquired hyperspectral images mixed with a variety of different types of noise. Therefore, HSI denoising is a necessary and very important process.Recently, a variety of hyperspectral image denoising method have been proposed, these methods can effectively remove specific noise, but, most of the methods can not simultaneously remove various types of noise. However, low-rank representation(LRR) based methods have been used for HSI denoising, which can simultaneously remove different types of noise: Gaussian noise, impulse noise, dead lines and so on. However, the LRR based method cannot make full use of the spatial information in HSI. In the HSI paper, we integrate the superpixel segmentation into the LRR, which can obtain the proposed SP-LRR. In detail, first, we use the principle component analysis(PCA) to obtain the first principle component of HSI. Second, we adopt the superpixel segmentation method on the first principle component of HSI to get homogeneous regions. Since we take advantage of both the spectral and spatial information of HSI by combining PCA with superpixel segmentation, thus it is better than simply dividing the HSI into square patches. Third, we employ the LRR to each homogeneous region of HSI, which can simultaneously remove the above mentioned mixed noise.Finally, experiments on both simulated and real hyperspectral images demonstrate that the proposed SP-LRR is efficient for HSI denoising.