Image Watermarking Algorithms with Synchronization Mechanism Using Self-Synchronization Decoding and Feature Points

In this paper we propose two image watermarking algorithms, which are robust to cropping operation using self-synchronization mechanism. In both algorithms, the watermark sequence is firstly encoded by non-cyclic Low Density Parity Check Codes (LDPC) and then it is embedded into the block-wised Discrete Cosine Transform (DCT) coefficients using the Quantization Index Modulation (QIM) algorithm. In the watermark extraction process, the first algorithm extracts the watermark sequence using self-synchronization of sum-product decoding. While the second one firstly extracts some synchronization code embedded previously in the regions centered by feature points of the image, and using them block segmentation starting position can be calculated. The desirable performance of both algorithms about the watermark imperceptibility and robustness against cropping attack with different cropping rate is shown, together with computer complexity.