Palm lines include principal lines and wrinkles, which can uniquely describe an individual palmprint. In this paper, a palm-line extraction approach is presented. It contains two stages: global filtering and local filtering. In the first stage, steerable filters are used in a whole image to extract rough palm lines and obtain approximate direction angles. In the second stage, using steerable filters further extracts palm lines and connects broken palm lines in local regions of a same palmprint image. Experimental results demonstrate the performance effectiveness of this approach.