A robust approach to enhancement of multivariate images

This paper addresses the problem of attenuating noise from vector-valued image data. A class of nonlinear filters stemming from robust estimation is introduced. An exact algorithm requires extensive computation. Therefore, approximate algorithms for computing the filter output are developed. The filters produce reliable results even if the assumptions on noise process are only approximately true. The performance of the techniques is studied using simulated data and data from range imaging sensor and in the case of additive multivariate noise with equal component variances, unequal component variances, correlated noise components, and in the presence of outliers.<
>