SIFT matching method based on support description

To reduce the image matching errors caused by local structure similarity and other factors, a matching judgment method based on support description was proposed. An initial matching set was obtained by scale invariant feature transform(SIFT) algorithm, from which the more stable feature points were extracted to build a support feature set. According to the distribution of support feature points, a support description on the remaining feature points of the initial matching set was performed. And similarity degree between the generated descriptors was used to determine whether the feature points match correctly. After judgment the correct matching feature points were added to the support feature set, so that the support feature set expanded dynamically and distribution density of the support feature points and accuracy of the support description would be guaranteed. Experimental results show that the proposed method can preserve the correct matches while eliminating more than 90% mismatches and improve the correct matching rate effectively.