Computationally efficient rotation invariant discrete cosine transform-based semi-blind watermarking technique

摘要

Digital image watermarking is very widely used for ensuring authentication and security of data transmitted over a network. Most of the watermarking techniques have been designed to accomplish robustness against various geometric and photometric transformations. The traditional method of embedding watermark using Discrete Cosine Transform (DCT) coefficients is more robust against noise and JPEG compression attack as compared to Zernike Moments (ZMs) based watermarking techniques. But the drawback of the DCT-based watermarking is that it is not rotation invariant. With DCT-based watermarking technique, a slight rotation on the watermarked image distorts the watermarked information completely. In this paper, ZM-based rotation angle estimation technique has been used to make DCT-based watermarking rotation invariant. Its performance has been compared with ZM-based watermarking. It has been shown through experimental analysis that in addition to its inherent robustness against noise and JPEG compression, the proposed watermarking technique is faster and more robust to rotation transformation as compared to ZM-based watermarking techniques.