Digital watermarking based on the scalable part of H.264 in MPEG-4-Scalable Video Code (SVC) is a promising technology in the field of video watermarking. However, the existing watermarking algorithms for H.264/SVC framework are difficult to satisfy the requirement of copyright protection. In this work, we extend the existed framework for invisible watermarking adopting the Discrete Cosine Transform (DCT) with a scalable watermarking algorithm, in order to adapt the multi-resolution property of the H.264/SVC. The original watermark is re-constructed to different scales by the pyramid decomposition, and the re-constructed watermarks are embedded in the DCT sub-blocks by modulating their positive and negative coefficients. Additionally, the sub-blocks are adaptively selected from the macro-blocks by a designed key control mechanism. According to the test results, the embedded watermarks are invisible and robust. The proposed method has achieved the Peak Signal to Noise Ratio (PSNR) in a value no worse than 41 dB and the Normalized Correlation (NC) in a value greater than 0.85. The proposed algorithm has can resist the attacks with different quantization steps.