A new fragile watermarking approach for tamper detection and recovery of document images

In this paper, a fragile watermarking scheme is proposed for tamper detection and recovery of digital images. In this scheme, the source document image is divided into blocks of uniform size and a pseudo-random mapping of blocks is generated. For each block, a watermark consisting of authentication and recovery information is generated and embedded into the corresponding mapping block. The authentication data is formed using the Least Significant Bit (LSB) of each pixel value in the block. The recovery information consists of sub-sampled versions of two moment-preserving thresholds of the block. At the receiver side, the watermark is extracted from the corresponding mapping blocks for each block of the watermarked document image. A dual-option parity check method is applied on the authentication data in the extracted watermark to check whether the watermarked image block is tampered or not. If the image block is found to be tampered, the recovery information is extracted from the watermark. From the experimental results it is found that the average value of tamper detection rate and recovery rate for various types of intentional attacks on the document images corpus is 93.8% and 93.6% respectively. Further, the average Peak Signal to Noise Ratio (PSNR) of the recovered images is 61.24%. The comparative analysis reveals that the proposed scheme outperforms the existing system [16].