The high cost resulting from a large number of mutants hinders mutation testing in practical application. In order to greatly reduce mutants under weak mutation testing, a new approach to reducing mutants based on statistical dominance analysis is presented. In the proposed approach, mutant branches are first constructed by combining statements before and after mutation, and a new program is formed by integrating all mutant branches into the original program. Furthermore, the dominance relations among mutant branches in the new program are determined by statistical information of coverage after executing test cases. Finally, the non-dominated mutant branches are obtained corresponding to the mutants after reduction. The proposed approach is applied to test eight programs, and the experimental results demonstrate that it can reduce 90% mutants on average, therefore greatly improve the efficiency of mutation testing.