Noncentral image moments for invariant pattern recognition
Lixin Shen · Optical Engineering · 1995
When invariant moments are used as numerical features for image description and recognition, the noncentral moments generated by correlation can be used for scale, rotation, and shift-invariant pattern recognition. A search in the feature-vector space for matching the input and reference feature vectors can be fast. The new method can apply to multiple-image input without segmentation of the images, and the noncentral moments are more robust than the central moments to noise.