Robust Blind Video Watermarking by Constructing Spread-Spectrum Matrix
Fei Zhang, Hongxia Wang, Ling Yang, Mingze He · 2022
Video watermarking is a quite important means to trace the distribution history of the video content by embedding the owner’s ID or some fingerprints (i.e., watermark) into each video frame. Therefore, it is hoped that the watermark of each frame can be extracted completely and correctly for video traceability and authentication. However, most existing video watermarking schemes are unavailable to extract the watermark information from each frame completely. Considering these limitations, we present a robust video watermarking method by constructing spread-spectrum matrix. Our idea is to constructing the spread-spectrum matrix based on the knowledge that human eyes are less sensitive to changes on high frequency component than low frequency component in order to improve watermark imperceptibility. Besides, we use the auto-convolution function (ACNF) to determine the location and scope of the embedded watermark and standard Hough transform (SHT) is used to determine the parameters of geometric distortion to improve watermark robustness. By evaluating various videos, the proposed video watermarking method guarantees good imperceptibility and a high watermark extraction accuracy when suffering from attacks such as video encoder compression, rotation, scaling, cropping, frame rate conversion and screenshot.