Dual-Plane Correlation-Based Video Watermarking for Immunity to Rotation, Scale, Translation, and Random Distortion

Abstract

A robust video watermarking method is described that can embed watermarks immune to not only rotation, scaling, and clipping and translation but also random geometric distortion and any of their combinations. It can detect watermarks without any search for canceling the effect of random distortion. Rotation, scale, and translation are canceled by searching for the angle, size, and origin of the original picture. The search for the angle and size is combinatorial and independent of the search for the origin. That is, the angle and size of the target picture can be restored without necessarily restoring the picture’s origin. Furthermore, the search for the origin is divided into horizontal and vertical searches that can be carried out independently. The number of searches is thus drastically reduced, making the processing time short enough for practical application. Watermark strength is controlled using an improved human visual system model for processing color information. This is done in the L*u*v* space, where human-perceived degradation of picture quality can be measured in terms of Euclidean distance. The watermarks are actually embedded and detected in the YUV space, where watermarks can be detected more reliably than that in the L*u*v* space. Experimental evaluation using actual video samples demonstrated that the proposed method can embed watermarks immune to rotation, scaling, clipping and translation, random distortion, and combinations thereof while maintaining picture quality.