Efficient Data Hiding with High Payload

Good rate-distortion performance is desired in data hiding, and the theoretical bounds are dependent on the range of allowable modifications made on cover samples. This work proposes a novel data-hiding scheme suitable for various modification ranges, and a family of data-hiding methods with different embedding rates and distortion levels can be generated. In the scheme, the secret digits are represented as a summation of vectors derived from cover samples in a shift manner. By introducing a trellis construction, the distortions of the possible sample-value sequences that can represent the secret data are calculated and only the sequence with the lowest distortion is selected as a new cover. It is shown the rate-distortion performance approaches the theoretical bounds with high embedding rate.