Resizing compressed images is often needed for different display, storage and transmission. Based on the analysis of energy reserving capacity for different basis vectors, we propose a two-layer downsizing scheme, where the elementary layer represents the low frequency components of the image, while some high frequency details can be recovered from the enhancement layer. The new method provides a solution for low definition image scalable transmission, at the same time, provides a potential tool to rationally recover the high definition information. Experimental results show that the new scheme exhibits better picture quality than some current downsizing approaches, with a comparable computational cost.