Peer Group Filtering of Noisy Color Images

In this paper a new filtering scheme for impulsive noise reduction in color images is proposed. The novel method is based on the evaluation of the statistical properties of a sorted sequence of aggregated distances used for the estimation of the vector median of samples from the filtering window. The filter is based on the Fisher's linear discriminant, which enables the detection of outliers injected into the image by the by the noise process. The new filtering design enables the extension of the peer group filtering concept into color image denoising.