Model Limitation and Matching Coefficients (MLMC): A new algorithm for feature matching

-In most algorithms for feature matching, people firstly construct a similarity measure function, then find the optimal solution using iterative methods, along with which the optimal matching matrix and transformation model parameters are gained. The selection of the initial iterative value is a key of the whole algorithm, because it influences the speed of the convergence and determines whether the algorithm can converge to the real value or not. This paper proposes an algorithm combining model limitation and similarity information for computing initial iterative value. The introduction of model limitation and new version for computing matching coefficients can overcome the problem of mismatching caused by the same or similar objects and ensure to get a good initial value. Experiment results show that the proposed algorithm is effective and potent. Index Terms Image registration, feature matching, model limitation, similarity information, matching coefficient.