Feature Transform of Global Invariant Moments in SIFT Algorithm

This paper present a SIFT descriptor that combines global geometric in this paper. The algorithm takes the feature point as the center, carries out the log polar coordinate transformation and establishes the concentric circle region, By analyzing the transformation of HU moments in discrete state, extracting the Feature vector of the invariant of the concentric circles in the discrete state with RSTC invariance. Tacking the Euclidean distance as a metric for image matching experiments. Experiments show that the algorithm which combines the global with local feature improves the robustness of image matching under various changes such as rotation and scale changes, viewpoint changes, illumination changes and image blur.