An Accelerated Mean Shift Algorithm Based on Prediction Model

In order to reduce computational load of mean shift algorithm,an accelerated mean shift algorithm based on prediction model is proposed.Convergence prediction model is built according to the different convergence properties of the iterative sequences,and acceleration is realized by reducing the distance between the vector and convergence point at each iteration.It is proved theoretically that convergence speed of our algorithm is quicker than the original one.The experimental results also show that the convergence speed is improved greatly and the tracking accuracy is guaranteed.