Illumination Robust Mean Shift Tracking

王路,阳琳谮,卓晴,王文渊

DOI: https://doi.org/10.3724/sp.j.1087.2008.01672

2008-01-01

Abstract:A new illumination robust Mean Shift tracking method was proposed.This method combined color feature and local binary pattern feature to describe the target.The illumination invariant characteristic of local binary pattern made the model more robust.To avoid the unstable bin-to-bin similarity applied in the original Mean Shift tracking method,a cross-bin similarity was used to make the feature matching more stable.Experiments show that the proposed method is robust than the original Mean Shift tracking method under illumination changes.