The region of interest detection for remote sensing images is usually based on global research and setting up the basis of prior knowledge. The new method called salient region detection based on salient features clusting for remote sensing images is proposed. We use the color information to construct the histograms in different color channel (RGB) to compute the information maps in each color channel. After fusing the information maps, we can get the single saliency maps. To get the saliency maps in CIELab color space, we adopt the k-means to cluster all the images in the CIELab color space, which makes it possible to reduce the computational complexity by calculating saliency on cluster-level. Then, through studying the integration of single saliency map and CIELab saliency maps, we get the final saliency maps. Finally, we can construct the mask of region of interest according to the final saliency map, which enable us to get the region of interest segmentation. Result shows that compared with existing models, we get more accurate saliency maps without the basis of prior knowledge. This method will be meaningful in further remote sensing image processing.