In order to improve the camera calibration accuracy, characteristic points extraction method with high accuracy is needed. Three typical calibration patterns including checkerboard pattern, crossed sine fringe pattern, and Gaussian point matrix pattern are chosen. Corresponding characteristic points extraction algorithms are analyzed and compared, adding the comparison of accuracy in camera intrinsic parameters calibration, the best one is crossed sine fringe pattern based on Fourier phase analysis. To avoid the drawback that the Fourier method used in the crossed sine fringe pattern cannot handle with characteristic points located in the edge of the point matrix, a novel method that using phase-shift method instead of Fourier method is proposed to obtain the crossed phase. The new method can eliminate the indistinct affect in the edge of the point matrix caused by Fourier translation filtering, so to amplify the scope of application of crossed sine fringe pattern in camera calibration. The extrinsic parameters calibration experiment show that the new method is effective in handling point matrix including edge points, for the reprojection error keeps in original level.