Point correspondence by matching scaled invariants

This paper investigates the point correspondence problem of two sets using scaled transform invariants. We proposed a method based on the idea of scaled invariant vector. A unit vector representation of invariants and scaling constant is introduced for affine transform. The method in principle applies to any dimension and any transformation with an computable invariant. The proposed method is shown to compute useful correspondence even with large number of outliers in noiseless case. We have also demonstrated improved performance comparing to most existing approaches for solving point pattern matching problem. The method is efficient and robust to noise up to a certain level.