The two-point algorithm is a linear correction method based on calibration data. Because of a narrow dynamic range and a poor accuracy, it is difficult for the algorithm to overcome the non-linearity and spatio-temporal drift of a detector in response. To remove the nonuniformity of the detector, a new algorithm which combines a polynomial fitting algorithm with a one-point algorithm is proposed after the two-point algorithm is analyzed. With this new algorithm, the variation of the coefficient of the detector with the environment is removed.