High-Dimensional Statistical Distance for Object Tracking
Yang Zhang, Ye Shufan, Yang Xiang, Liqun Gao · 2010
This paper deals with object tracking in the video sequences. The goal is to determine in successive frames the object which best matches. So we used the similar measure between the reference object and candidate object can be distinguished: Relying on the same principle of histogram distance, but within a probabilistic framework, we introduce a new tracking technique. First, measure based solely radiometry include distances between probability density function of color histograms. Then we propose to compute the Chebyshev distance between high-dimensional PDF without explicitly estimating the PDF. The distance is expressed directly from the sample using the nearest neighbor framework. It capability of the tracker to target object variations, is demonstrated for several image sequences.