A Restorative Image Authentication Scheme with Discrimination of Tampers on Image or Watermark

Most image authentication schemes in the literature are either vulnerable to vector quantization (VQ) counterfeiting attacks or sacrifice localization accuracy to improve the security. Only a few algorithms exist to take the recovery of the tampered areas into account. This motivates us to develop a novel authentication scheme for the accurate tamper localization and restoration of the tampered areas. In particular, the host image will firstly be divided into blocks, the recovery watermark of each block is embedded into another distant block selected by a key, and the localization watermark of each block is embedded into itself. During generating the localization watermarks, the sliding window is used to put resistance against VQ attack. It is the dual watermarks that can distinguish tampering watermark from that of image. Theoretical analysis and experimental results demonstrate that the watermarked image has higher quality, and what's more the position of tampered areas can be located correctly. Additionally, it can restore the tampered region effectively.