DOI: 10.3724/SP.J.1004.2013.02071

Acta Automatica Sinica (自动化学报) 2013/39:12 PP.2071-2076

Alternating Direction Method for Salt-and-pepper Denoising

Conventional image denoising algorithms attempt to remove noise on the basis of spectral separation between signal and noise. However, in practice, the spectra of signal and noise usually overlap. As a result, conventional denoising algorithms often suppress noises at the cost of losing details in fine texture and causing blurring in output images. This paper formulates the image denoising problem by adopting a new model of sparse and low-rank matrix decomposition. Based on this model, the alternating direction method (ADM) is utilized to obtain the restored image. Our experiment demonstrates that the ADM suppresses salt-and-pepper noise more efficiently and better preserves detail information in the input image, as compared to the commonly used median filtering method.

Key words:Image denoising, convex optimization, l1-norm, nuclear norm, alternating direction method (ADM)

ReleaseDate:2014-07-21 17:04:34

Funds:National Natural Science Foundation of China (61102097, 61102096), Science and Technology Support Program of Tianjin (11ZCGHHZ00700), and Natural Science Foundation of Tianjin (11JCYBJC06900)