L. Kotoulas and I. Andreadis (Greece)
Pattern Recognition, Image moments, Image reconstruc tion
Image moments are extensively used in various image anal ysis tasks. Moments of a discrete orthogonal basis were recently proposed as image descriptors, presenting some advantages over moments of a continuous basis. In this pa per, we derive some of their properties regarding geomet rical transformations, as well as some computational char acteristics. Most of these properties hold for all orthogonal polynomial moments, but when applied to discrete ones, no approximation errors appear. By using these properties, we propose a novel method for the efficient computation of the moments of a 2D object through its projections.
Important Links:
Go Back
