Fast computation of 2-D image moments using biaxial transform

In this paper a new transform is introduced. This transform is based on div iding the image into four quadrants. The transform can be generated by addi ng or subtracting the gray levels of each two axially symmetric points in t he original image. The addition or subtraction is dependent on the quadrant position. The newly transformed quadrants can then be used to compute the moments. Only one quadrant of transformed image is needed to compute the mo ments of specific orders. This can lead to considerable savings in both pro cessing time as well as storage. Other advantages of this transform, such a s image reconstruction and using multiple filters on the same image frame t o remove noise, are also explored. (C) 2001 Pattern Recognition Society. Pu blished by Elsevier Science Ltd. All rights reserved.