Rewritable data insertion in encrypted JPEG using coefficient prediction method

This paper proposes a data insertion method using DCT coefficient prediction algorithm in encrypted JPEG image. First, a sequence of quantized AC coefficients are removed, and the remaining nonzero coefficients are shifted to the left, hence ensuring sufficient room for data insertion. Next, the external data is mapped to a set of Huffman codewords, and the external data is inserted into the image by introducing the corresponding Huffman codewords. At the receiver's side, the hidden data can be extracted by referring to the pre-defined table without the need to decrypt the image. To restore the image, the decrypted JPEG image is segmented using the Superpixel function, and further processed using the prediction algorithm for recovering the removed AC coefficients. Experimental results show the recovered images resemble their original counterparts. In the best case scenario, up to 19% compression gain is achieved while 24, 576 bits of external data can be inserted into the image.