基于Markov随机场的运动目标快速分割.pdfVIP

28 7 Vol. 28 No. 7 2007 7 M ICROCOMPUTER APPLICATIONS Jul. 2007 M arkov 陈文楷 张利洲 张青竹 曾 涛 崔 刚 ( 100022) : MRF, , EM , MRF, , ICM : EM ICM Segm entation ofM oving Object Based onM arkov Random Field C EN W enkai, Z ANG Lizhou, Z ANG Qingzu, ZENG Tao, CU IGang (College ofElectronic Information ControlEngineering, Beijing University ofTechnology, Beijing, 100022, China) Abstract: One self- adapted segmentationmethod formoving object is applied in this chapter. W euseM ixedGaussM odelto describe the difference of serial mi ages and estmi ate parameters ofGaussM odelby adoptingAnnealing andAdapted EM algorithm. M arkov Ran- dom F ieldM odel is established on the basis of that estmi ation. Energy functions are constructed according to M arkov Random F ield M odel. Image is expressed by a configurationwhen the total energy function value ismaxmi ized by adopting ICM A lgorithm. K eywords: Segmentation ofM oving Object, Annealing andAdapted EM A lgorithm, M arkov Random F ield, ICM A lgorithm , [ 1] : ,,,(); , , (MRF: M arkov Random Field) [ 2, 3, 4] G G S = { 1, , L } , (),, ,,;MRF Gibbs, , , Gibbs, Gibbs 2 2. [ 5] 1: {: x G} -MRF, Gx ,x x 2005-10-10 674 2007 , x G 2:() S = { 1, , L} P = p ( l) , ( l S ), -() Gibbs, Gibbs G ibbs 1 -U( l) p ( l) = e ( 1) Z U( l) -, Z , Z = e-U(l) Gibbs l S ammersley- Clifford: , -MRF -Gib