Multispectral Transforms Using Convolution Neural Networks for Remote Sensing Multispectral Image Compression

A multispectral image is a three-order tensor since it is a three-dimensional matrix, i.e.one spectral dimension and two spatial position dimensions. Multispectral image compression canbe achieved by means of the advantages of tensor decomposition (TD), such as NonnegativeTucker Decomposition (NTD). Unfortunately, the TD suffers from high calculation complexity andcannot be used in the on-board low-complexity case (e.g., multispectral cameras) that the hardwareresources and power are limited. Here, we propose a low-complexity compression approach formultispectral images based on convolution neural networks (CNNs) with NTD. We construct anew spectral transform using CNNs, where the CNNs are able to transform the three-dimensionspectral tensor from large-scale to a small-scale version. The NTD resources only allocate thesmall-scale three-dimension tensor to improve calculation efficiency. We obtain the optimizedsmall-scale spectral tensor by the minimization of original and reconstructed three-dimensionspectral tensor in self-learning CNNs. Then, the NTD is applied to the optimized three-dimensionspectral tensor in the DCT domain to obtain the high compression performance. We experimentallyconfirmed the proposed method on multispectral images. Compared to the case that the newspectral tensor transform with CNNs is not applied to the original three-dimension spectral tensorat the same compression bit-rates, the reconstructed image quality could be improved. Comparedwith the full NTD-based method, the computation efficiency was obviously improved with only asmall sacrifices of PSNR without affecting the quality of images.