The work aims to design an adaptive video watermarking algorithm based on pixel evaluation and mo-tion compensated prediction error expansion in order to solve these defects as embedding distortion and low watermark capacity induced by ignorance of the correlation between the video pixel characteristics and the time in current video watermarking technology. Firstly, the motion compensated prediction error expansion was introduced to predict the gray values of each video frame and obtain the corresponding motion compensated prediction error. Then, the local standard deviation of pixel motion compensation was calculated according to the prediction error, and the whole video was divided into smooth and non-smooth sub-blocks. A pixel suitable for embedding watermark was selected from the video according to the local standard deviation of motion compensation and the pixel selection threshold. The watermark embedding mechanism was designed for different pixels based on the watermark capacity and the chosen video pixel to hide the watermark information into the video specific pixels. Finally, the watermark extraction method was constructed to recover the initial watermark information from the watermark image. The test data showed that this algorithm had higher watermark capacity and lower distortion when compared with the current video watermarking method. The maximum watermark capacity was 2 bits and the PSNR value of watermark information was more than 40 dB. This algorithm has good imperceptibility and low distortion, so it has a certain application value in copyright protection, information security and other fields.