This paper proposed a DCT domain blind watermarking algorithm. Firstly we divide the original image into 8*8 blocks and conduct discrete cosine transform to these blocks, then extract the transformed DC coefficient of each block to form a coefficient matrix, and then best-scramble the coefficient matrix to compare with the chaos-processed binary watermarking image, to achieve the objective of watermark embedding by quantifying the DC coefficients of host image according to the result of comparison. The watermarking extraction process does not require the original image so that achieved the blind extraction. Simulation shows that the algorithm has good robustness and transparency in cope with common attacks.