Proposal of Nonlinear Image Denoising Algorithm for Images Corrupted with Gaussian and Impulse Noise

가우시안과 임펄스 잡음이 혼재한 이미지에 적용하기 위한 비선형 잡음제거 알고리즘의 제안

The statistics for the Gaussian noise mixed with impulsive noise are modelled. The denoising algorithm called amplitude-limited sample average filter is derived, which is optimal in terms of minimizing mean square errors under the assumption that contaminating noise is heavy-tailed Gaussian distributed. Its performance is shown to be excellent when image is corrupted mainly with Gaussian noise. However, it shows visually grainy output as the amount of impulsive noise increases. In order to overcome such problems, it is combined with the myriad filter to propose an amplitude-limited myriad filter. Simulation shows it effectively removes both Gaussian and impulsive noise, not blurring edges severey.