An in-focus lens position may be determined by computing a focus metric value, by counting the number of transition pixels for images of a target captured at different lens positions. Using as little as two frames to compute two focus metric values, a reasonable approximation of the in-focus lens position may be obtained. The approximation of the in-focus lens position may then be used as a starting point for a fine focus search process, to determine an in-focus lens position. An advantage here is that the focus metric values relate to the number of transition pixels and are easy to compute, and yield a reasonable approximation of the in-focus position with just a few frames. Other embodiments are also described.