Iterative method of movement estimation for television signals

An iterative method for accurate movement estimation of television signals is analyzed. The method considered is based on an iterative process applied to the conventional gradient method. It is known that the gradient method has some problems both in its stability and in its accuracy for movements larger than 2 or 3 pel/frame. It is shown that when an iterative process is used, the estimated movement converges to the actual movement rapidly and is exponentially stable. It is also shown that the asymptotic accuracy of the iterative estimate is determined by the smoothness of the gradient and by the amount of the frame-to-frame noise. Some experimental results are given and compared to the theoretical result