Blind IR spectral deconvolution for image feature extraction via sparse representation regularization

Image feature extraction is a significant task for the computer to understand the natural environment. Infrared imaging spectrum is an efficient tool to achieve this. However, the infrared spectroscopic data often exists the problems of random noises and peaks overlap. In this paper, we propose a blind infrared spectral deconvolution with sparse representation regularization for image feature extraction. The Bandelet transform is applied on the degraded IR spectrum and clean spectrum, their distributions of the Bandelet transforms coefficients are compared. By incorporating the sparse prior to the spectral deconvolution process, we could effectively constrain the inverse problem and control the ringing artifact which is often produced by the traditional algorithms. Compared with serval state-of-the-art methods, the proposed method shows its effectiveness in the ability of noise suppression and overlap band splitting. The recovered infrared spectrum data can be used for extracting accurate image feature in COREL dataset.