Two-step non-local means method for image denoising

• Published in: Multidimensional systems and signal processing Vol. 33; no. 2; pp. 341 - 366
• Main Author: Zhang, Xiaobo
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.06.2022; Springer Nature B.V
• Subjects: Artificial Intelligence; Circuits and Systems; Electrical Engineering; Engineering; Feature extraction; Image processing; Iterative methods; Noise; Noise reduction; Signal,Image and Speech Processing; Smoothness; Wiener filtering; Non-local means (NLM); Wiener filter; Center weight (CW); Denoising amount; Image denoising
• ISSN: 0923-6082; 1573-0824
• DOI: 10.1007/s11045-021-00802-y

Non-local means (NLM) method is a powerful technique in the field of image processing. The center weight CW (the weight of the pixel to be denoised) plays an important role for the performance of NLM. In this paper, several center weights such as Zero-CW and One-CW are researched in the influence of these weights on denoising performance. In order to avoid the disadvantages of excessive smoothness or insufficient denoising of these different NLM filters, a two-step non-local means (TSNLM) iterative scheme is proposed. In the first step, local Wiener filter is introduced to extract image features from the method noise of NLM with Zero-CW. The denoising process is integrated into NLM based on local Wiener filter (LWF-NLM). In the second step, the carefully selected NLM (NLM with One-CW) operates on the output of the first step to remove the remaining noise. The denoising amount of two steps is combined by the decaying parameter depending on noise variance. As far as I know, this is the first time to consider the role of center weight to design an iterative NLM filter. The experimental results show that the proposed TSNLM helps NLM to improve the ability of denoising, giving satisfactory subjective and objective performance. Furthermore, the proposed TSNLM is very efficient compared to other related NLM based iterative methods.