Salt-and-Pepper Noise Elimination by PI-Remover and Modified AMF Noise Detectors

This paper proposes a new filtering scheme _ PI remover for effectively eliminating the salt-and-pepper noise of images. The PI-remover can be created straightforwardly based on the concept of proportional-integral (PI) controller. Similar as the design of PI-controller, the PI-remover also has the proportional parameter k_P and integral parameter k_I, that are used to weight the noisy pixels. The modified adaptive-median filter (AMF) is used to detect the possible noisy pixels. The noisy pixels are compensated to approach an objective value by running few iterations of the PI-remover´s compensation loop. The experimental results demonstrate that the PSNR of the restored images by using PI-remover is significantly improved as compared with that by using AMF. The PI-remover can still perform well as the noise level increases to 90%.