In this paper, a new approach for calculating star camera attitudes, which can calculate calibration parameters and attitude information simultaneously, in situations where precise camera calibration results are unknown, is proposed. This algorithm combines the calibration and attitude determination processes, achieving significantly improved performance as a result. Experiments using 1500 star images from different sky regions show that, compared with the traditional method that separates calibration and attitude determination, the proposed algorithm not only exhibits more precise and stable results, but also shows improved tolerance for the star mismatching which is inevitable in star sensor data process technology.