In the photographic measurement with high precision, the image processing precision plays a crucial role in the overall measurement accuracy. In the process of imaging, the discrete sampling of the feature points by cameras causes a distortion between the received signal and the original image, which brings errors in image processing. By analyzing the total energy, the dispersion radius of energy distribution of the feature points and the image processing windows, and based on the relation of the feature point center position, extraction error and the standard deviation of the energy density function, an error compensation method is proposed for discrete sampling. According to the method, only once calibration of the compensation parameters is needed for any camera and algorithm, and the image processing precision can be improved significantly. The experiments show that for the centroid method and the Gaussian fitting method, the image processing precision can be increased to 0.03 pixel.