The concept of preprocessing is widely used in various computer science area such as compiler and software engineering for the purpose of macro processing and optimization. In addition, preprocessing is also used in SAT solvers in order to eliminate redundant literals and clauses to speed up its solving time before searching the state space. However, there is an unexpected run-time error such as stack-overflow during this step, in case the size of a given set of clauses is huge which impedes SAT solvers. In this case, the preprocessing should be applied at the encoding time to optimize its size. In this paper this idea is applied to several Sudoku problems. As a result, significant improvements are obtained with respect to the number of variables and clauses as well as the solving time compared to the previous works.