Vector directional order-statistics for impulse detection

In this paper, a new vector directional approach for impulse noise detection and filtering in color images is provided. The novelty of this approach lies in the use of vector order-statistics achieved by the ordering according to the sum of angles between input vector samples, in other words directional information, as the base for impulse detector. Thus, the input samples are separated into two classes such as noise-free samples and corrupted samples. This simple binary decision is performed by the comparison between the operation value and the angle threshold. An importance of the vector directional order-statistics with the smallest angle distances to input samples consists in the determination of the operation value that is given by an angle between the central sample and the mean of `smallest' vector order-statistics. After the separation, only affected samples are processed by basic vector directional filter, whereas noise-free samples are passed to a filter output without change, i.e. the system performs an identity operation. In order to achieve the optimal performance of the proposed method, there is necessary to determine the threshold angle and the number of considered vector order-statistics, too. Since, the proposed adaptive method represents directional processing, the excellent results in the term of the simultaneous impulse noise suppression and signal-details and color chromaticity preservation are desired.