A new method based on quadratic programming theory is proposed to determine the regularization parameter. A function whose variable is the derivative with respect to the regularization parameter is formulated. Based on the property that the value of the function is very dependent on the regularization parameter, the optimal parameters can be determined by using quadratic programming theory. A numerical example is utilized to select the parameters through the method. It is concluded that the new method can effectively overcome the ill-posed problems from an ill-conditioned system matrix, and that it can obtain an approximate solution with higher accuracy, fine stability and effective noise resistance.