A space-frequency Non-Local means image denoising algorithm

Un algorithme de débruitage Non-Local means espace-fréquence

Simon Postec (1) , Jacques Froment (1) , Béatrice Vedel (1)

1 LMBA - Laboratoire de Mathématiques de Bretagne Atlantique

The efficiency of the Non-Local means (NLM) image denoising algorithm relies on the identification of similar original pixels from noisy similar patches. Hence fine details and low-contrasted structures are badly recovered after the application of NLM. But as these structures tend to correspond to redundant ones in the Fourier domain, NLM filtering in this domain allows one to better denoise them. A mixed space-frequency approach improves the denoising performances of NLM because it ensures that the information is redundant enough, in the spatial domain or in the frequency domain. Our approach is simple : it consists in running two times the NLM algorithm (firstly in the frequency domain and secondly in the spatial domain). For fine textures and isolated points we get a better visual reconstruction than with the original NLM. In terms of PSNR, the improvement can be over 1 dB. Our approach gives intermediate results between the original NLM and state-of-the-art methods while at the same time having moderate complexity and leading to few visual artifacts.

Domaines: Analyse fonctionnelle [math.FA]; Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Simon Postec, Jacques Froment, Béatrice Vedel. Un algorithme de débruitage Non-Local means espace-fréquence. XXIVe Colloque GRETSI - Traitement du Signal et des Images, Sep 2013, Brest, France. pp.ID181. ⟨hal-00904761⟩