An Adaptive Random-valued Impulse Noise Reduction Method Based on Noise Ratio Estimation in Highly Corrupted Images

In this paper, we propose a novel random-valued impulse-noise reduction method by adaptive edge-preserved median filtering, called AEPMF, with adaptive threshold and noise-ratio estimation. Generally, a pixel is always very similar to its horizontal and vertical neighbors and thus such a characteristic can be used to estimate the noise-ratio of a corrupted image for deriving appropriate thresholds. To substantially reduce noises, AEPMF uses iterative PSNR-checking strategy in which if the PSNR of the previously filtered image is lower than a threshold, next filtering process is executed. Experimental results manifest that the proposed AEPMF method is more robust and effective than other switching-based median filters and achieves above 20% in average PSNR improvement rate when the corruption ratio is above 30%.