光照鲁棒的Mean Shift跟踪方法

针对Mean Shift跟踪方法存在的光照不稳定问题，提出了一种光照鲁棒的Mean Shift跟踪方法。该方法采用颜色特征和局部二元模式特征（Local Binary Pattern)两种特征相结合来描述目标，其中局部二元模式特征的光照不变性使得目标模型更加鲁棒。同时，为了避免原始Mean Shift跟踪方法中bin-to-bin度量带来的不稳定性，该方法采用了一种新的cross-bin度量，该度量更好地融合了多层次的特征信息，使得光照变化下的特征匹配更加稳定。实验表明，该方法在光照变化情况下能取得比原始Mean Shift跟踪方法更好的性能。

Illumination robust Mean Shift tracking

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.