关键词: 特征点密度, 数字水印, 几何攻击, 离散余弦变换, 盲检测

Abstract: To solve the problem that the digital watermarking images suffer geometric attacks, a strong robust blind detection digital watermarking algorithm, integrated with the unchanged features of the image and the characteristics of the frequency-domain stability, was proposed. By constructing the self-adaptive Scale Invariant Feature Transform (SIFT) algorithm and filling the seats with the Harris method of the angle and the point, utilizing the density of feature points obtained in different scale space and adjusting the watermarking information self-adaptively to the intensity embedded into Discrete Cosine Transform (DCT) domain, it scrambled the image sub-block feature vector set with Arnold scrambling, and then generated the key documents, next matched the two-way feature with the secret image feature vector, got geometric distortion parameters and corrected the image restoratively, and extracted the IDCT inverse transform domain watermark information in the form of blind detection. Judging from the experimental results, the algorithmic Peak Signal to Noise Ratio (PSNR) in this paper improves by 13% than using Discrete Wavelet Transform (DWT) and Discrete Fourier Transform (DFT), and the similarity of watermark is over 11%. It shows that this algorithm has good robustness to both of the geometric attacks and the conventional signal processing, as well as better invisibility.

Key words: feature point density, digital watermarking, geometric attack, Discrete Cosine Transform (DCT), blind detection