Infrared spectrum resolution enhancement model via Gabor transform regularization for object detection

Abstract

In this paper, we proposed a novel resolution enhancement model for the unknown instrument function and infrared spectral signal estimation. The Gabor transform can be utilized on the degraded spectrum to save the proposed resolution enhancement model from converging the mistake spectral signal. In the proposed model, the spatial random noises are regularized by the total variation regularization, as well as the spectral local smoothness is reduced by the Gabor coefficient distribution regularization. Then, the split Bregman iteration scheme is introduced to alternately optimize the instrument function and latent spectrum, which shown excellent convergence and stability. After those steps, the high-resolution spectral signals can be produced by our proposed model. Comparing with the existing resolution enhancement algorithms, the proposed model can produce the comparable quality even if the observed spectrum exists the low contrast.