A Complete Set of Pseudo-Zernike Moment Invariants by Image Shape Description

摘要

Orthogonal moments such as pseudo-Zernike moments have been successfully used in the field of image analysis. Conventionally, image function is mapped onto a set of orthogonal functions over the unit circle. If the origin of polar coordinate system is taken at the centroid, the rotation invariants will be easy to obtain. Based on pseudo-Zernike moments, this paper presents a new method to drive the complete rotation, scaling and translation (RST) invariants from the orthogonal projection transform (OPT). The efficiency and the robustness to different noises of the method for classification tasks are presented by comparing it with several existing methods.