The work aims to propose a method of detecting the surface defects of the mobile phone screen glass based on the machine vision technology to solve the problems of low accuracy and poor stability in the manual detection of the surface defects of the mobile phone screen glass. The statistical average method was used to establish the template image to reduce the influence of the external illumination on the gray level of the template image. Firstly, the pixel alignment between the template image and the image to be tested was implemented by the registration method based on mutual information. Secondly, the residual image was obtained by subtracting the template image from the aligned image to be tested. Finally, the Niblack method was used to determine the defects of residual image. 300 images of mobile phone screen glass were obtained through the experimental platform established and their defects were detected by the proposed method, the template matching method and the manual method. The experimental results showed that, the proposed method achieved the true positive rate of 92%, the true negative rate of 96.5% and the accuracy of 95%. Compared with the template matching method and the manual detection method, the proposed method improved at least 5%, 4% and 4.3% in terms of true positive rate, true negative rate and accuracy respectively. The proposed method outperforms the manual detection method in terms of accuracy and stability of surface defect detection of mobile phone screen glass.