Abstract:

Most feature-based image registration algorithms only use one kind of feature, and each feature is considered as of equal importance during the process of finding features correspondences. These limit the range of applicability and impact the robustness of the algorithm. To solve this problem, a new algorithm using multi-feature is proposed to register images on the base of defining the uncertainty of features. The usage of the  uncertainty makes sure that the robust features are more influential to the similarity function. In order to compute the maximum value of the multi-feature similarity measure function, two methods using alternate strategy and constructing uniform matrixes are proposed respectively. Experiments show that the proposed algorithm is effective and potent.