Multispectral image compression method and system based on spatial-spectral feature separation and extraction

Abstract

The invention discloses a multispectral image compression method and system based on spatial-spectral feature separation and extraction. Spectral features and spatial features of a multispectral image are extracted through an inter-spectral feature extraction module and a spatial feature extraction module respectively; pixel-by-pixel fusion is carried out, downsampling is carried out in a unified mode, and the size of a feature map is reduced; A compressed data code stream is acquired by quantization and entropy coding; when the image is recovered, entropy decoding and inverse quantization are carried out on a received code stream to obtain characteristic data, the characteristic data is sent to a reverse decoding network, the structure of the reverse decoding network is symmetrical with that of a forward coding network, and the space and spectrum information of the image is recovered through an up-sampling recovery size and a corresponding space-spectrum characteristic recovery network. A reconstructed multispectral image is acquired. According to the invention, through the spatial-spectral feature extraction module, spatial and inter-spectral information is separated and extracted, the integrity of spatial-spectral features is ensured, and high-performance multi-code-rate compression of a multi-spectral image can be realized in combination with a compression network and rate distortion optimization.

Description

technical field [0001] The invention belongs to the technical field of image processing and deep learning, and in particular relates to a multi-spectral image compression method and system based on spatial spectral feature separation and extraction. Background technique [0002] Remote sensors can generate three-dimensional multi-spectral images containing rich spectral and spatial information by collecting digital images of multiple continuous narrow-band spectral bands. Marine applications and other aspects have a wide range of applications. However, with the rapid development of multispectral imaging technology, the spatial resolution of multispectral data is getting higher and higher, resulting in a rapid increase in its data volume, which is not conducive to image transmission, storage and application, and will hinder the further development of related technologies. Therefore, it is of great practical significance to study efficient image compression techniques. [00...

AI Technical Summary

Problems solved by technology

There are obvious deficiencies in these three types of algorithms: Although the algorithm based on predictive coding can achieve lossless compression, the compression rate is low, and the design of the predictor is the main factor affecting the compression performance of the method. A good predictive algorithm can Significantly reduce the entropy value of the residual image; the algorithm based on vector quantization technology is extremely complex, and since the degree of distortion is inversely proportional to the size of the codebook, in order to reduce the degree of distortion, the capacity of the codebook must be greatly increased. When the codebook is too large, the amount of calculation will be greatly increased; when the compression rate of the algorithm based on transform coding is large, there will be block effect and edge Gibbs effect, which will seriously affect the compression performance

The convolutional neural network is the most widely used nowadays, but the general convolutional network usually ignores the spectral information of the multispectral image, which will cause a large amount of information loss.

Images