Adaptive embedding and detection for improved video watermarking

Video watermarks must maintain picture quality and withstand video processing. Resolving the tradeoff between these conflicting requirements has been one of the central problems of video watermarking. Previous video watermarking methods, however, have trouble satisfying these conflicting quality and survivability requirements because they utilize the watermarking methods for still picture and neglect the properties of motion pictures. To resolve this tradeoff, we describe adaptive video embedding and detection techniques that utilize the properties of motion pictures. Motion-adaptive embedding can allocate watermarks to picture areas which are adaptively based on motion properties of video and, statistically adaptive detection can control the accumulation of watermarks based on the statistical properties of video to prevent degradation in the watermark signal by video processing. Experimental evaluations using actual motion pictures have shown that these techniques can be effective for resolving the tradeoff and can be widely used in the pixel-based watermarking.