In the existing automatic focusing systems, the traditional evaluation functions are sensitive to noise and have small focusing range which will affect the accuracy of the auto-focusing results. An autofocus evaluation function based on gradient operator with variable frequency was proposed. It solved the problem that the coverage of the evaluation functions is small when the fuzzy degree is larger by sampling frequency reducing method in the image space. As mentioned above, an absolute value gradient operator with threshold value selection was adopted, which not only improved the noise resistance but also reduced the complexity. Then, the simulation experiments and actual tests were carried out to validate the satisfactory performance of the method. The experimental results indicate that it can meet the characteristics of wide focusing range and high sensitivity at the same time when the gradient operator with different sampling frequencies was used and then compared with the existing traditional methods. The quantitative indicators were used to illustrate that the calculation speed of the evaluation function were improved a lot compared with the existing evaluation function. The autofocus evaluation function proposed was demonstrated has better anti-noise performance and has variable focusing range with high sensitivity and high real-time, which can evaluate the sharpness degree of the images in the process of automatic focus accurately.