An Algorithm of Adaptive Deformation Estimation of Moving Object in the Mean Shift Algorithm
Jifeng Ning, Shuqin Yang, Fuzeng Yang · 2009
An algorithm of adaptive deformation estimation of moving object in mean-shift tracking method is developed. Firstly, the differences between object and background in the weight image associated with the target candidate region are analyzed. By their different distributions, the area estimation of the target is converted into the task of image segmentation. The threshold is automatically selected by maximizing the between-class variance and then object and background are automatically segmented. It leads to the scale of the object estimated. Finally, the width, height and orientation of the object are estimated by combining the estimated area and covariance matrix. For three video sequences, the experimental results validate its robustness to the deformable estimation of the targets.