Automatic focusing is very important to high-quality image acquisition for camera system. This paper proposes a new automatic focusing method using skin-color face detection to automatically determine focusing area. In order to overcome the influence of color temperature on the face detection method, we propose a new color correction method by learning a color temperature curve with statistical methods. The color temperature curve is a concise representation of the inverse relation between light sources in the logarithmic color space. The method compensates color shifts of images by locating a set of compensation gains from the trained color temperature regression curve. A comparative setup of experiments is built to verify the effectiveness of the proposed method. Controlled experimental factors include light source, object distance and initial lens position. Modulation transfer function of images is applied to objectively evaluate the image quality before and after the new automatic focusing method. Experimental results show our method has high success rate of face localization, and effectively achieves automatic focusing.