P3-1 ABayesian Network-Based Deinterlacing Scheme
Gwanggil Jeon, Hyojoon Choi, Dong‐Hyung Kim, Joohyun Lee · 2008
Thispaperproposes a novelmotionadaptive interpolation algorithm forvideodeinterlacing. We employa Bayesian network, whichprovides accurate weightings among commonexisting deinterlacing methods. Itsuccessively builds approximations ofthedeinterlaced sequenceby weighting interpolation methods. Theresults ofcomputer simulations show thattheproposed methodoutperforms anumberofmethodsin theliterature. I.INTRODUCTION Themotion adaptive (MA)deinterlacing method isakind ofspatio-temporal interpolation methods, whichrequire varying levels ofcomputation power, andprovide varying levels ofimagequality. Thechallenge inimplementing an effective MA algorithm ismanaging thetrade-off between double images andlost vertical resolution. Another challenge invarying thetemporal aperture isavoiding resolution pumping asobjects start orstopmoving. Thispaper proposes thenewMA deinterlacing algorithm whichemploys Bayesian Network (BN)andthree deinterlacing methods.