This paper addresses the problem of designing recursive L-filters optimized
by the least mean square (LMS) algorithm. The sorting operation brings in
nonlinear behaviour into the L-filter structure, and thus it hampers the de
rivation of a closed-form solution for optimizing the weights suited to rec
ursive LMS L-filters. An alternative training strategy is proposed to itera
tively derive the weighting coefficients for recursive L-filters. Simulatio
ns conducted show the advantage of the recursive L-filter over its nonrecur
sive counterpart in suppressing image noise. (C) 2001 Elsevier Science B.V.
All rights reserved.