JPEG is a versatile and widely used format for images. Based an elegant design that enables the joint works of basis transformation (gross-scale decorrelation) and entropy coding (fine-scale coding), the resulting JPEG image can maintain virtually all visible features of an image while reducing its size to one tens of the original raw data. Motivated by the increasing amount of images taken from the same or similar scenes and stored in different resolutions, an interesting question aries: can JPEGs that originate from similar (but not identical) image source be jointly compressed - better than independently compressed? Addressing this important question essentially requires extracting hidden correlations that exist between JPEGs from “similar” original raw images, which is the main contribution of this paper. In fact, by developing an “invert-and-project” (IVP) prediction approach, we are able to utilize one of the JPEGs to compress another one at coarser resolution, obtaining a compression ratio that is over 65% better than the state-of-art lossless compression methods specialized for JPEG data.