Modify Additive Lee's Filter for Filtering images Corrupted by Film Grain Noise

Abstract

In m:rny intagc- proccssing :rpplic:rtions the noise corrttpts theinr:rge ii signal dcpcndcnt, t5c nrost n'idely cttcounterccl typcsbeing nrultiplicativc suclt ils lilnr grain noisc. Ttt1i1 contnrolrfeature is that the poN'cr of the noise is rcl:rted to thc brightncss ofthe corrupted pixel. This results in brighter ilreils :rppelring to benoisier than darkcr areils. In tliis pilper,.lvc introduce modifyaclditivc noise Lec's filter for filtcring inrages corruptcd by signaldepenclent film-gr:rin noise.This by assuming that the noise variilncc not constant over thervhole image regions, ancl can be estimatcd localll'. Results ofMddify Lee's filter are conlpared w'ith the multiplicatil'e versionsLee's fitter standard, both in terms of subjective and objectivemeasures. The modify Lee's filter provides better noisesuppression as inclicatecl by lower normalize mean square error inhomogenous image regions and edge regions than the stanclarclLee's filter.