A model for detecting salient regions in an image based on location shift and extent trace is proposed in this paper. In this model, the detection is divided into two stages: the first one is location shift at which the location of a new salient region is found according to the global saliency measurement, and the second is extent trace at which the size of the salient region is chosen according to the local saliency measurement. In this way, a series of salient regions are detected through the alternation of location shift and extent trace. Based on this model, a novel algorithm is presented. The results of experiments with many real images show the algorithm is effective and efficient.