Infrared image reconstruction based on a modified block compressed sensing

Abstract

Regarding to the reconstruction problem of an infrared image in compressed sensing, we proposed a modified block compressed sensing method. First, the original infrared image is divided into small blocks, each of which is sampled with a Gaussian random matrix to generate a small amount of measurement data. Second, Spectral Graph Wavelet transform with excellent sparse features is applied into reconstruction process of projected Landweber algorithm, and the hybrid median filter is used for enhancing image smoothness and reducing block artifacts. Finally, the high-quality infrared image satisfied termination conditions is obtained. Experimental results on various types of infrared images show that the proposed method attains much better performance in CS recovery than the conventional ones and can obtain higher quality infrared images.