Efficient Convolutional Neural Network for Spectral-Spatial Hyperspectral Denoising

Denoising is a common pre-processing step prior to analysis and interpretation tasks such as classification, unmixing and target detection, typically carried out for hyperspectral images (HSIs). In this paper we develop which performs spectralspatial HSI denoising through a convolutional neural network (CNN). Our newly developed method, called single denoising CNN (HSI-SDeCNN), considers HSIs as 3D data cubes, performing the denoising process with only one single model. Experimental results on both synthetic and real data demonstrate that our newly developed HSI-SDeCNN outperforms other stateof-the-art HSI denoising methods.