Hyperspectral Image Enhancement And Dimensionality Reduction

Hyperspectral images have been exploited to observe our world for which they record the spectrum of scene radiance by a distribution of intensity at each wavelength.The availability of such a physical representation of the scene plays a vital role in numerous applications.However,it is difficult to achieve higher classification accuracy while preserving the relevant original information of the hyperspectral images.First,the hyperspectral images acquired by sensors are inevitably disturbed by noise,which may lead to inaccurate results.Moreover,the low spatial resolution of HSI is incapable of capturing the details of the scene or land cover.Finally,the enormous spectral bands of HSIs provide large volume and redundant information,further cause a big increase in the computational time and make the analysis of hyperspectral images become a fairly complicated task.The contributions of the this paper can be summarized as follows:(1)To remove noise in hyperspectral image,a spectral–spatial kernel method for hyperspectral image denoising is proposed.The proposed method is inspired by the observation that the spectral–spatial information is highly redundant in HSIs,which is sufficient to estimate the clear images.A spectral–spatial kernel regularization is proposed to maintain the spectral correlations in spectral dimension and to match the original structure between two spatial dimensions.(2)We propose a new framework to enhance the resolution of hyperspectral images by exploiting the knowledge from natural images: The relationship between low/high-resolution images is the same as that between low/high-resolution hyperspectral images.In the proposed frame-work,the mapping between low-and high-resolution images can be learned by deep convolutional neural network and be transferred to hyperspectral image by borrowing the idea of transfer learning.(3)To remove the highly correlated bands,a multigraph determinantal point process(MDPP)model is proposed to capture the full structure between different bands and efficiently find the optimal band subset in extensive hyperspectral applications.(4)To reduce the computational burden,a different dimensionality reduction framework is proposed to perform feature extraction on the selected bands.The proposed method uses determinantal point process to select the representative bands and to preserve the relevant original information of the spectral bands.The performance of classification is further improved by performing multiple Laplacian eigenmaps on the selected bands.