Adaptive vector LUM smoother
R. Lukac, Stanislav Marchevský · 2002
A new vector approach, the adaptive vector LUM smoother for color images is presented. The novelty of the proposed method lies in the two-stages adaptive choice of the estimate, where in the first filter stage (N+1)/2 outputs with all smoothing levels done by the vector LUM smoother with window size N are generated. In the next filter stage, one of (N+1)/2 possible samples is adaptively selected to determine the filter output depending on the local information, by the filter window. Thus, the proposed method provides an optimal estimate in the sense of objective criteria. In addition, the filter output is constrained to be a sample from an input set that cannot result in color artifacts.