SUSAN controlled decay parameter adaption for non‐local means image denoising

Abstract

The non-local means (NLM) method explores self-similarities in images for noise removal. The traditional NLM (TNLM) method computes pixel similarity using the globally fixed decay parameter. However, a fixed decay parameter for the whole image makes it difficult to ensure that the TNLM method can restore both edge pixels and non-edge pixels effectively. To address this problem, the SUSAN controlled NLM (SNLM) method is proposed to adaptively tune the decay parameter for each image pixel based on the SUSAN edge response. Extensive simulations demonstrate that the SNLM method outperforms several state-of-the-art NLM methods in terms of noise reduction and detail preservation.