Based on the background continuity priori knowledge, we propose a new saliency-detection method. In this method, an image is first abstracted as a set of super-pixels. We then search the shortest path and calculate the foreground weight of each super-pixel in the path according to its color difference with those at two ends of the path. We also integrate the color-difference-based foreground weight correction with the boundary priori knowledge based saliency optimization to obtain the final saliency detection results. The proposed method is extensively evaluated on several public datasets and the experimental results confirm its state-of-the-art performance.