Markov Random Field Models in Computer Vision-英文文献.pdfVIP

Markov Random Field Mo dels in Computer Vision S Z Li Scho ol of Electrical and Electronic Engineering Nanyang Technological University Singap ore szlintuacsg Abstract A variety of computer vision problems can b e optimally p osed as Bayesian lab eling in which the solution of a problem is de ned as the maximum a p osteriori MAP probability estimate of the true lab eling The p osterior probability is usually derived from a prior mo del and a likeliho o d mo del The latter relates to how data is observed and is problem domain dep endent The former dep ends on how various prior constraints are expressed Markov Random Field Mo dels MRF theory is a to ol to enco de contextual constraints into the prior proba bility This pap er presents a unied approach for MRF mo deling in low and high level computer vision The unication is made p ossible due to a recent advance in MRF mo deling for high level ob ject recognition Such unication provides a systematic approach for vision mo deling based on sound mathematical principles Intro duction Since its b eginning in early s computer vision research has b een evolving from heuristic design of algorithms to systematic investigation of approaches for solving vision problems In their search for solutions researchers have real ized the imp ortance of contextual information in image understanding In this pro cess a variety of vision mo dels using context have b een prop osed Among these are Markov Random Field MRF theory based mo dels of which analytic regularization theory based mo dels are sp ecial cases MRF mo deling is app ealing for the following reas