Due to the diversity of geographical objects, it makes great sense to introduce multi-scale segmentation/representation into the analysis and interpretation of high spatial resolution remote sensing images. However, with the increasing use of highresolution images, traditional multi-scale segmentation methods gradually show their shortages in efficiency, especially whenhandling large-scale images. In this paper, a novel bi-level scale-sets model (BSM) is proposed for multi-scale region-based representation of large-scale remote sensing images. In BSM, first, an image is divided into blocks with overlapped margins, and a low-level scale-sets model is block-wisely implemented. Second, a segmentation result is obtained by retrieving and mosaicing the block-wise segmentation results, based on which a high-level scale-sets model is implemented covering the whole image. To further improve the efficiency of BSM, a parallel implementation is presented for the block-wise scale-sets model. In the experiments, first, the effectiveness of the BSM is validated using a worldview2 image covering a coastal area of Shenzhen, where the BSM obtains accurate multi-scale representation results without any mosaic artifacts. Then, the efficiency of the BSM is demonstrated by comparing with the state-of-the-art multiscale segmentation method, the one integrated in the commercial software eCognition v9.2, where the proposed BSM takes about 7 minutes to process a 24000×24000 multi-spectral ZY3 image, and is 2-3 times faster than the competing method.