A Neural LUM Smoother
Stanislav Marchevský, Rastislav Lukàč · Brno University of Technology Digital Library (Brno University of Technology) · 2000
In this paper a design of neural LUM smoother is presented. The LUMsmoother distinguishes by a number of smoothing characteristics done bythe filter parameter. However, the tuning parameter for smoothing isfixed for whole image. The new method realizes adaptive control of thelevel of smoothing by neural networks. The well-known and very popularbackpropagation algorithm is used. The analysis of the proposed methodsis evaluated through subjective and objective criteria and comparedwith the traditional LUM smoother.