| In this paper, I get the feature points by using the Harris corner, SIFT features and the application of improved Hausdorff distance, I proposed the Hausdorff distance image registration based on the Harris and SIFT(Scale-invariant feature transform)features to achieve the image registration. The Hausdorff distance has stronger anti-noise ability and rotating robustness, it makes us to distinguish these features points are whether right or not when we use the Hausdorff distance to match feature points, even if there are many error points we can ensure the correctness of matches are not subject to interference from the feature points, At the first, we use Harris corner detection and SIFT feature to extract the corner points of reference image and the registered image, we found that the searching range of corner points was extended after using these two methods, at the same time, the error points were eliminated by using similarity matching principle, to further improve the accuracy of the image feature points by increasing the number of the right feature points while reducing errors number of feature points,. To minimize the defect which the Hausdorff distance need large amount of calculation when there are huge number of points, In this way we can greatly reduce the computational complexity and improve the accuracy of image registration. This process of the image registration has a stronger anti-noise ability and rotating robustness. This paper proved that the matching feasibility through the experimental test from different image sources or under different conditions and the scale and angle of deformation in the case of image registration results are very consistent with experimental error requirements.The experimental results demonstrated that the running time of improved algorithm declined45%when compared with traditional Hausdorff distance algorithm, and have stronger anti-noise ability and rotating robustness, which improve the efficiency and accuracy of registration. |
PDF Full Text Request