The autofocus of the scenes containing high bright areas such as light sources is common and the high brightness areas lead to saturation of the image pixels. However, the existing evaluation functions are put forward for normal scenario and these cannot be applied for the scenes with light sources such as the nightscop. A sharpness evaluation function is proposed to remove saturate pixels by analyzing the performance of scenes including light sources in the process of defocus and it can be used to evaluate the image quality of scenes containing light sources or high bright areas. A binary map of focusing window is acquired by modifying the Ostu algorithm, and an expansion region of high bright areas is obtained by morphological expansion method to prevent the dispersion of bright areas caused by defocus blur. The removed bright areas template is put on the gradient information matrix, and the average gradient is used as sharpness evaluation value index. The sharpness evaluation index is tested using multiple sets of focal sequence diagram. The results demonstrate that the proposed autofocus evaluation function is applicable to the scenes containing light sources or high bright areas compared with other four common evaluation functions. The proposed method has favorable unbiasedness, unimodality, stability and broad coverage.