Impulse noise removal: Noise detection versus pixel estimation

Abstract

This paper proposes a new finding regarding the different roles and significance of noise detection and pixel estimation utilized in the majority of impulse noise removing algorithms. Extensive experimental results will be presented to show that the importance of each method depends on the noise ratio. In other words, by particularly examining how the adaptive median filter (AMF) algorithm [18] employs the two aforementioned methods in comparison with their modified versions as well as the ideal-filtering algorithm, this paper will point out the role that each method plays. The significance of noise detection can be realized at low noise ratio while pixel estimation gains its efficiency as noise ratio increases.