Применение преобразования Уолша в системе видеосопровождения для вычисления корреляции изображений объектов

Abstract

The paper deals with the problem of increasing the object tracking system accuracy by reducing the delay in the tracked object coordinates determination block. The computationally complex object images correlation is computed when determining coordinates. For computational complexity reduction it is proposed to use fast Walsh transform (FWT), by which a dyadic correlation function is computed, witch, afterwards, is converted to cyclic using sparse factored matrices. Due to low FWT computational complexity and matrix factorization, the proposed correlation computation method has better performance comparing to existing methods.