Abstract
The continuous template matching method have been proposed in the previous papers in order to realize the precision subpixel displacement detection between subsequent frames of motion pictures. In the proposed method, a successive approximation algorithm adopting partial derivatives was used for maximizing the interpolated cross correlation functions. The authors suggested that using partial derivatives leads to some weakness in the search of the maxima. In this paper, a nueral network based maximization (peak detection) algorithm is proposed in order to resolve the above mentioned fragileness.
