ABSTRACT
One of the classical approaches for information retrieval and signal characterization is the use of invariant moments [1, 2] that has been widely applied for pattern recognition and computer vision systems. The invariant moments are measurements of a signal that are invariant to some transformations such as rotation, scaling, translation or illumination. Moments can be applied to many different aspects of image processing, ranging from invariant pattern recognition and image encoding to pose estimation. When applied to images, they describe the image features with respect to its axes. They are designed to capture both global and detailed information of the image. In its continuous formulation, an image can be considered as a two-dimensional Cartesian density distribution function f(x,y). With this assumption, the general expression of a moment of order (p+q) evaluated over the complete image plane X; is:
where iff is the weighted kernel or basis function. This moment produces a weighted description of f(x,y) over the entire plane xj. The basis functions may have a range of useful properties that may be passed onto the moments, producing descriptions which can be invariant under rotation, scale, translation and orientation.
