Hyperspectral intrinsic image decomposition based on local sparseness

Abstract

Due to the influence of conditions such as sensor status, imaging mechanism, climate and light, hyperspectral remote sensing images have serious distortion, which seriously affects the classification accuracy of hyperspectral remote sensing images. In this paper, the intrinsic image decomposition technology, which is widely used in computer vision and graphics, is introduced into hyperspectral image processing to perform intrinsic image decomposition on the original image. A hyperspectral intrinsic image decomposition method based on local sparseness is proposed. The automatic subspace partition and sparse representation theory are used to decompose the original hyperspectral image. The reflectance intrinsic image obtained by the decomposition is subjected to hyperspectral image classification processing. The experimental results show that the method proposed in this paper can obtain the intrinsic images better, and improve the classification accuracy of hyperspectral images to a large extent.