DCT-Based Video Watermarking Transcoding Technique

This paper presents a new technique of how to insert additively imperceptual watermarking to resist downsampling attack according to Watson´s model based DCT domain and the block classification. Firstly, we obtain subimages through quasi-convoluted down-sampling technique. Secondly, downsampling video block is classified as five categories by computing the luminance block variance. Finally, the region and strength of optimizing watermark complied with standard normal distribution (zero mean and unit variance) is additively embedded into the unmarked original down-sampling image by exploiting the luminance characteristic and a perceptual distortion constraint, which greatly improves robustness of upsampling watermarked video. The original video is not required for watermark detection. Simulation results in video quantity degrading averagely with PNSR 3.4 dB but subjective quantity remaining. The proposed scheme is not only robust to downsampling attack, but also achieves a low computational complexity.