Subpixel edge location based on orthogonal Fourier–Mellin moments

Abstract

In order to satisfy the stringent requirements for edge location accuracies in such fields as medical image analysis and satellite remote sensing, an approach based on orthogonal Fourier–Mellin moments is proposed to use the lower radial orders and the rotation invariance of these moments to describe small objects in images, and edge location is accomplished by setting the edge in the vertical direction and analyzing the interrelationships among the orthogonal Fourier–Mellin moments with different orders and degrees, and consequently the specific characteristics of the edge can be fully extracted. Experimental results show that the proposed method can be used to achieve an edge location accuracy of 0.16 pixel for straight lines with noise and 0.23 pixel for curves with noise. It can therefore be concluded that the proposed method is an efficient approach to satisfy the stringent requirements for higher edge location accuracies in such fields as medical image analysis and satellite remote sensing.