Local/Non-local Means for Multi-Modal Denoising

By effectively leveraging image periodicity during the denoising process, the non-local means filter has demonstrated superior performance over some conventional methods. However, its limited consideration of local image information may lead to the removal of useful non-periodic details during denoising. To address this limitation, this paper introduces a novel filter that integrates both local and non-local means. Furthermore, given the increasing prevalence of multimodal images in real-world applications, the proposed filter is extended to handle multimodal scenarios. Results from classic image experiments, including both subjective visual assessments and objective quantitative metrics, indicate that the proposed filter surpasses the original algorithm in denoising performance.