Image Matching Based on Local Entropy and Pyramid Transform

In this paper we determine matching positions of key points based on criteria of local entropy difference. Local entropy of each point in images is computed, and sequences of local entropy are analyzed using pyramid transform. We adopt pyramidal data structure, and find crude matching point in lower resolution image, and determine accurate matching point in original image ( higher resolution image) finally. These techniques have led to raising computing effect. This image matching method has very good anti-noise properties and anti-distortion of geometry properties. The experiment results show that matching accuracy is up to 78% when intensity of salt & pepper noise is under 5%.