Gaussian-Geometric Moments and its Application in Feature Matching

Since the 7 famous Hus invariants had been introduced in 1960s, the moment invariants play an important role in image analysis and pattern recognition. In this paper, we propose a new moment called Gaussian-Geometric moment, and derived their translation and rotation invariants. One significant conclusion drawn is that the rotation invariants of Gaussian-Geometric moments have the identical forms to those of geometric moments.The Gaussian-Geometric moments and the geometric moments both can represent the image information, difference is that the Gaussian-Geometric moment can represent the center information of an image and the geometric moments represent the edge information of an imageonly. This is particularly evident in the performance of high order ones. Another important property of Gaussian-Geometric moments is that it has a scale parameter which allows choosing the best scale to represent the interest region of an image. A detailed comparison has been made to test the feature matching capability between the proposed moments and the geometric moments. The results show that the proposed moments perform much better than the geometric ones.