Vector order-statistics for impulse detection in noisy color image sequences

This paper focuses on a new three-dimensional vector filtering approach for the impulse noise suppression in noisy color image sequences. Like the vector median-based filters, the proposed method utilises a reduced ordering scheme. However, in order to approach the optimal filtering situation, when only affected samples are estimated whereas noisy image points are passed to a filter output without the change, the proposed method is controlled by the binary decision rule. As a switch between the identity filter (no filtering operation) and the vector median serves a simple comparison of the filter threshold and Euclidean distance between the central sample and the mean of several vector order-statistics with the smallest distances to the input samples. The proposed method is characterised by excellent results, especially in terms of the mean absolute error criteria.