A blind video watermarking method was proposed,in which the watermark composed of symbol part and in-formation part was embedded into the original video frames. After some video frames hiding the watermark were selected randomly from one segment of the video sequences,the symbol water mark was embedded into low frequency discrete cosine transform (DCT) coefficients of fixed blocks of such frames,whereas the information water mark was embedded into direct current coefficients of some other blocks selected by a cryptographic key. Repeat embedding till the whole video was watermarked. To extract the information watermark,the frames holding watermark were found out by detect-ing water the symbol mark rapidly from one video segment whose length was long enough to hold the whole watermark. Then,by using the high correlations between the watermarked frame and its neighboring frames,the pixel values of the blocks where the watermark was embedded can be well estimated. Comparing the estimated values with the current val-ues in the corresponding positions,the information watermark could be reconstructed. Experimental results show that this video watermarking method is robust against attacks such as compression,scaling,frame dropping,etc.
