An anti-noise subpixel algorithm of phase-shifting of fundamental frequency was presented based on the phase-shifting of Fourier transform and the anti-noise characteristics of low-frequency part of the phase spectrum of the image. The essence of the algorithm is that the displacement caculation of the image is replaced by the movement caculation of the coordinate, which makes the phase of the fundamental frequency zero under different coordninates when image position changes. Under the circumstances that the image of the CCD autocollimator is polluted by the noises caused by tempreture, the measuring accuracies of the normally-used barycenter, edge detection, Gaussian fitting algorithm and the algorithm presented in this paper were compared. Experiment results show, the subpixel algorithm demonstrated here has the advantages of strong anti-noise ability and high precision. The reliability of the algorithm is also disproved by the peak location of the reconstructed image after the removal of higher harmonics. When applied to the one-dimensional CCD photoelectric autocaollimator used in field conditions, fine linearity and ±3// measurement accuracy were simutaneously obtained in the whole ±3600// measurement range when the temperature varies between -400C-600C.