Abstract: Saliency detection is to find the most important object automatically according to the human visual habit in unknown scenes. In order to further improve the effectiveness of saliency detection and reduce the computational complexity of the pixelbased detection algorithm, this paper proposes a saliency detection method based on dynamic guided filtering. In the novel simple iterative guided filtering, the kernel function uses the input image and dynamic image to substitute the fixed guided image like the classic guided filtering. This ensures the better structure transmission from the input image to the guided image. Secondly, in order to save the time cost of the algorithm, the sampling method is used to reduce the amount of calculation needed. Finally, in order to extract more effective saliency regions, a key saliency region extraction method is introduced. It can obtain the more accurate object region by modifying the key points set. The experimental results show that the proposed algorithm is more quick and more effective compared with other pixelbased saliency detection methods.