Recoverable Video Watermark in DCT Domain

Few of current video watermarking algorithms can prevent HD videos from unauthorized copying after the production is distributed. This paper proposes a recoverable video watermarking algorithm in DCT domain to achieve this goal. Watermark is first split and modulated by CDMA. Then, after applying boundary detection, frames in one shot are divided into blocks. Finally, watermark is embedded into DCT coefficients of these blocks. The embedded watermark could resist to conventional video processing such as loss compression, resolution change, and format change, etc. The security of embedded watermark is based on the key used in CDMA modulation. With correct user keys, authorized user can completely remove the watermark and restored the high quality video. The experimental results show that the proposed algorithm is robust and practical.