RGB image spectral information reconstruction method based on dictionary atom embedding

Abstract

The invention discloses an RGB image spectral information reconstruction method based on dictionary atom embedding, and belongs to the field of computer vision. According to the method, a hyperspectral image of a corresponding scene is obtained by reconstructing RGB image information, firstly, a training sample is obtained from a high-resolution spectral image library, and then a complete spectral dictionary is obtained through learning by using a Gaussian process based on a Bayesian nonparametric dictionary learning theory and considering the smoothness of spectral change; secondly, for each dictionary atom, selecting low-dimensional pixels and corresponding high-dimensional pixels according to feature similarity with the low-dimensional space, forming a mapping matrix from the low-dimensional space to the high-dimensional space by using neighborhoods of the low-dimensional pixels and the corresponding high-dimensional pixels, and completing a training stage; and finally, selecting a mapping matrix mapped from the low-dimensional space to the high-dimensional space according to the relationship between the test image and the spectrum dictionary, and reconstructing the mapping matrix, thereby completing the reconstruction of the hyperspectral image. According to the method, the high-resolution hyperspectral image can be reconstructed through the color image, meanwhile, the calculation speed is guaranteed, and the method can be applied to the fields of medical imaging, geological exploration, agricultural production and the like.

Description

technical field [0001] The invention belongs to the field of image processing, and relates to a method of using Gaussian process learning in consideration of the smoothness of spectral changes, obtaining a complete dictionary based on a Bayesian non-parametric analysis model, and reconstructing a high-spectral resolution image through a neighborhood embedding method. At the same time, it involves restoration A method for optimizing the reconstruction of spectral information by the camera spectral sensitivity function. Background technique [0002] Hyperspectral images are approximately continuous spectral curves of target objects acquired in a large number of electronic bands such as visible light, ultraviolet, and near-infrared, and reflect the spectral information and spatial relationship information of the spectral reflectance characteristics of the target object. The application of hyperspectral images has been extended to criminal detection, medical imaging, agricultura...

AI Technical Summary

Problems solved by technology

[0004] In view of the lack of interpretability of existing algorithms and the lack of adequate consideration of hyperspectral image features, etc.

Images