Abstract:

In the image registration based on feature, aiming at the inefficient of Hausdorff distance calculations based on the edge of images, a new method is proposed to calculate the Hausdorff distance in use of key point features that could show the space structure of images. This method is to reduce the quantity of matching point sets by detecting the odd point of images. The experimental results demonstrate the validity of this method, and it is superior in calculating and matching Hausdorff distance to that one based on edge. Aiming at the characters of sparse key point sets, an improved Hausdorff distance is proposed. This Hausdorff distance is more perfect for calculating the distance of sparse key point sets. The experiment shows this distance is better in matching the noised image than the other Hausdorff distance.