In this paper, a blind object-based video watermarking scheme, which is robust to MPEG4 compression and normal video editing such as Rotation, Scaling or Translation (RST), is proposed. The watermark is embedded into the Log-Polar Mapping (LPM) of DFT magnitude of the object; the synchronization problem introduced by RST is solved using PCA (Principle Component Analysis); by mapping the difference between watermarked video object and original video object from LPM domain to DFT domain, the watermarking can eventually be performed in the DFT domain so that video quality degradation caused by LPM and Inverse LPM (ILPM) is minimized. Experimental results verified the validity of the proposed robust watermarking scheme, which also has large watermark capacity.