Image matching through combined features of singular value and region

This paper proposes a new algorithm that combines the region feature and singular value vector in facing the instability when using the singular value vector separately in image matching to get a more accurate matching. The approximate singular value vector has been defined in this paper based on the fact that the singular value has a robust performance invariant to light the image disturbance and it is qualified to represent the image as an algebraic feature in general instance. A kind of region similarity of two binary images has also been defined in the algorithm. Experiments are performed on a large test set and the results demonstrate that the proposed algorithm has more veracious capabilities than the method using singular value vector separately.