An efficient object tracking method based on adaptive nonparametric approach
Longshuang Li, Zhixuan Feng · Opto-Electronics Review · 2005
In this paper, an efficient method for object tracking based on nonparametric approach is presented. The density we estimated is based on an adaptive kernel model, which is driven by the intensity difference between the target and the background. The background-weighted histogram for statistics of feature takes into account the relevance between the target and background. What is more, this approach extends the range that is needed for searching object. The target model is updated according to the change of the object and environment. Experimental results on real image sequences demonstrate its robust performance in visual tracking and require less iteration computations when compared to other method.