There are two kinds of auto-focus methods in a digital imaging system: Depth from focus (DFF) and Depth from defocus (DFD). Depth from defocus is a method, which acquired distance information from blurred image. Because of necessity of establishment of defocus model and accurate calibration of imaging system, the application of DFD is confined. As a search focusing technology, Depth from focus has some advantages, such as high accuracy, robust performance, so it is adopted in our apparatus. In this paper, an auto-focus apparatus based on digital signal processor has been designed and constructed. DSP receives image series from a CMOS sensor firstly, then performs the auto-focus action with the following steps: analyzing focus measure of image, moving motorized lens and finding the best imaging position. For focus measures, the frequency spectrum functions, the gradient functions and the entropy function are analyzed in detail. In order to improve the sensitivity of focus measure curve, the weighting factor - distance between direct current component and alternating current component - was introduced into revising frequency spectrum functions. A series of focus experiments have been performed on the auto-focus apparatus, and the veracity and the reliability of the system have been proved to be excellent.