A pixel frequency error compensation method of star sensor was introduced in detail, and the compensation effect was verified by the experimental data. At first, based on threshold segmentation algorithm for star extraction, the main reasons of pixel frequency error were analyzed. Then, the original point spread function of centroid location was improved and a pixel frequency error compensation method based on sub-pixel coordinates was proposed. Last, through the micro-pace experiment of star sensor, it was compared with sine curve method. Experimental results show that: in the center of the field of view, the pixel frequency error of the sampling point is reduced by 65.2% using this method, which is better than 52.7% of the sine curve method; using the error compensation formula of the field of view to compensate for the sampling point of the field of view, the pixel frequency error is reduced by 58.7%, which is better than 41.9% of the sine curve method. By the experimental results, compared with sine curve, this error correction method not only has better error compensation effect, but also has strong versatility within the field of view.