Abstract: A new fast algorithm called various mask correlation approach (VMCA) is proposed. In this method, correlation proceeds from sketch to detail, and a self-adaptive threshold is selected to remove some points that are not matched in each level so as to reduce the computational cost. Experiment results show that when the mask is bigger and the selected method of self-adaptive threshold is proper, computational cost of the fast algorithm is significantly reduced with a high precision and reliability at the same time, which is better than some existing fast algorithms.