A fusion approach for denoising impulse noise

Abstract

In the image handling process, effective interpretation or reconstruction of the original image is mandatory, since the post-processing of images relies upon this important pre-processing task. It is well recognised that the noise makes degradation in the quality of image, which in turn results poor interpretation of images. This paper targets to remove salt-and-pepper or fixed impulse noise from images by fusing two techniques, namely median and non-local means (MNLM) filtering, and exploits both filtering benefits altogether. In the first phase, median filtering has been applied along with new sorting algorithm instead of the conventional method of sorting in order to reduce computational and hardware complexity. Even though the non-linear median filter performs better for low-level noise density and best in edge preservation, it gives blurring effect in the case of high-level noise density. This paper tries to overcome this limitation by fusing NLM filter effectively during the second phase. This fusion approach has brought promising quantitative results such as higher PSNR (peak signal-to-noise ratio) and lower MSE (mean square error) when compared to other existing standard algorithms such as SMF (standard median filter) and NLM filtering approaches.

Keywords:

• salt-and-pepper noise
• standard median filter
• non-local means filter
• peak signal-to-noise ratio
• mean square error

Disclosure statement

No potential conflict of interest was reported by the authors.