Research on the algorithm of rapid object contour tracking

In this paper, color features and texture features are used as tracking clues, in a Bayesian framework, to estimate foreground probability and background probability, meanwhile, they are taken as their energy to construct curve theory functional and curve descending flow is used to optimize for the object of accurate contour. Compared with traditional tracking frame, LBP is used to describe pixel texture features in neighborhood of a curve in the algorithm proposed in this paper, and thus its computational complexity and process time are decreased, and also color instability is much improved in varying illumination.