Title: Infrared image denoising using a smoothed L₀ sparse regression

Authors: Shengqian Wang; Jianping Xiao; Chengzhi Deng

Addresses: School of Information Engineering, Nanchang Institute of Technology, Nanchang, Jiangxi Province, 330099, China ' School of Communication and Electronic, Jiangxi Science and Technology Normal University, Nanchang, Jiangxi Province, 330013, China ' School of Information Engineering, Nanchang Institute of Technology, Nanchang, Jiangxi Province, 330099, China

Abstract: Sparse representation based on image denoising has acquired considerable interest. In the most previous work on the sparse representation, L₁ and L₀ norm are always used as the sparsity regularisation. Despite the success of L₁ or L₀ norm, the limitation of this approach on its computational complexity or sparsity affects the efficiency or accuracy. In this paper, a smoothed L₀ approach based on infrared image denoising is proposed. Firstly, an improved smoothed L₀-based K-SVD (SL₀-KSVD) method for dictionary learning is presented. And then the infrared images are sparsely represented by the smoothed L₀ method and the coefficients are denoised using the constantly updated K-SVD dictionary. Finally, some experiments are taken on comparing the peak signal to noise ratio (PSNR) performance of our proposed method with its counterparts on different images. The experimental results on both the visualisation and real data of the infrared images demonstrate the superiorities of our proposed method.

Keywords: sparse representation; infrared images; image denoising; dictionary learning; K-SVD; peak SNR; signal to noise ratio; PSNR.

DOI: 10.1504/IJCSM.2015.069748

International Journal of Computing Science and Mathematics, 2015 Vol.6 No.3, pp.241 - 255

Received: 15 Jul 2014
Accepted: 27 Oct 2014
Published online: 08 Jun 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article