Hyperspectral image compression method based on spatial and spectral content importance

Abstract

The invention relates to a hyperspectral image compression method based on space and spectral content importance. The method comprises the following steps of: firstly, using a training data set for training a compression network model to obtain model training parameters, then dividing an input image tensor into two branches, subjecting one branch of image tensor to encoder network compression processing to obtain a 1 / 16-scale hidden representation tensor of an original image, inputting the hidden representation tensor into a quantizer network, performing pre-quantization and quantization processing to obtain a binarized code stream, inputting the other branch codes into a multi-depth convolutional network to generate an importance map, weighting the importance map and the binarized code stream to obtain a code stream based on content, and inputting the code stream based on the content into a decoder to obtain a reconstructed image. According to the method, the importance map is generated according to the spatial characteristics and the spectral characteristics of the hyperspectral image at the same time, and under the guidance of the importance map, the coding end can dynamically allocate the code rate according to the spatial and spectral content complexity of the image, so that the compression ratio is improved, and the image compression quality is ensured.

Description

technical field [0001] The invention belongs to the technical field of hyperspectral image compression, in particular to a hyperspectral image compression method based on the importance of space and spectral content. Background technique [0002] Compared with natural images, hyperspectral images not only have the redundancy of spatial correlation, but also the redundancy of similarity between spectral segments and between spectral segments. The rich spectral information of hyperspectral images can fully reflect the differences in the physical structure and chemical composition of samples, and can provide important data support for geological exploration, precision agriculture, and environmental testing. However, with the rapid improvement of the resolution of remote sensors, the data size of hyperspectral images increases geometrically, the correlation between the bands becomes stronger and the information redundancy becomes larger, which not only increases the computationa...

AI Technical Summary

Problems solved by technology

The above methods are all compression methods for natural images, not hyperspectral image compression

Images