Image blocks encoded by VQ are usually in the form of indices, which represent codewords' positions in a codebook C
. The VQ encoder E quantizes the input image block x
by selecting a best-matched code vector cb
}, where N
is the size of the codebook C
]. Here, the Euclidean distance given in Eq. (4.2)
is often employed as a metric for the best matching. It is obvious that this quantization might cause some distortion to the watermark information embedded in the pixel domain, if the embedding algorithms are sensitive to this operation, like most value expansion–based lossless information hiding algorithms [34
]. So a lossless watermarking algorithm that works ...