抄録
JPEG format, one of most commonly used image compression methods, focuses only on similarities of pixels in the same block. We present a new method to re-compress JPEG images by using similarities in multiple blocks decoded until the step before inverse DCT.