Vector quantization method capable of rapidly compressing high-spectrum signals

Abstract

A vector quantization scheme capable of rapidly compressing high-spectrum signals is provided for the problem that high-spectrum images are large in data size. Components of vector data are reordered according to dispersion, prior dimensions of components of the vector data are extracted to form a training vector set, the remaining components form remaining vector data, and clustering training is conducted on the training vector set; classification is directly conducted on the remaining vector data through code indexing of the training vector data; recombination is conducted on the code book of the training vector data and the code book of the remaining vector data to obtain the final code book of the vector data, and the compressed coding of the high-spectrum images is achieved. By means of the vector quantization scheme capable of rapidly compressing the high-spectrum signals, most of energy of the vector is made to be concentrated at the lower dimension part, only the lower-dimension-part vector data are extracted to be trained, and the space complexity is greatly lowered. The image restoration quality can be basically kept unchanged, the space complexity in the compressing process can be greatly lowered, various calculated amounts can be greatly reduced as well, and the target of high-spectrum image compression is achieved rapidly and effectively.

Description

Technical field [0001] The invention belongs to the field of hyperspectral remote sensing image processing, and specifically relates to a vector quantization scheme for quickly compressing hyperspectral signals. Background technique [0002] With the development of optical technology and sensing technology, in the 1980s, hyperspectral remote sensing technology emerged. This technology uses an imaging spectrometer with high spectral resolution for earth observations, and generates a set of three-dimensional hyperspectral images that combine spectral information and image information into one. Each spectral band of the hyperspectral image corresponds to a complete two-dimensional image of the observation scene, which is used to grasp the material space structure, and each pixel in the scene can extract a continuous spectral curve to determine the nature of the ground object. Therefore, the use of hyperspectral images can accurately achieve ground object classification, target dete...

AI Technical Summary

Problems solved by technology

[0008] Aiming at the existing hyperspectral image compression scheme, the problem of high computational complexity when obtaining higher image restoration quality, the present invention proposes a vector quantization scheme for rapidly compressing hyperspectral signals

Images