A new image matching algorithm for change detection using hilbert curve

Finding significant change in high resolution sensed image is an important task in maintaining GIS database. A class of these algorithms detects changed regions by means of edge comparisons. After extraction of feature points from a sensed image and a reference image, the feature points matching is a pivotal key in change detection. In general, given two point sets, find the minimum or maximal value of some measuring distances under the (affine) transformation. Because of the measurement errors and some outlying points, it is important that the measuring distances should be robust. Recently, a well known robust measuring distance called (partial) Hausdorff distance is widely used in feature points matching. It is more efficient than other conventional methods and has been applied in many fields. Although it is a reliable similarity measure, it is also a computational task. In this paper, we present a new algorithm using Hubert curve in order to resolve the computational complexity problem. This distance can be computed in the 1-D space rather than in the 2-D space that can greatly reduce the computational complexity. Our algorithm shows good performances for this task.