Invisible image watermarking is an effective method to provide protection for copyright ownership of digital media.Most image watermarking algorithms use pseudo random sequences as watermarks and so they can only give the answer of "Yes" or "No".Embedding meaningful watermark is important in some applications but results in more difficulties.There are two problems in embedding meaningful watermark as follows:(1) the embedded watermark must have a certain bit rate.And (2) the embedded watermark must be detected correctly from possibly distorted watermarked image.In this paper

we apply the theory and methods in digital communication to watermark embedding and propose an invisible image watermarking algorithm in DCT domain for embedding meaningful watermarks.The algorithm reduces error probabilities of detection by exploiting BCH code and matched filtering.With the proposed algorithm

we embed a meaningful watermark composed of 32 characters into an image of 256×256 and the watermark thus generated is robust.The watermark can be detected with zero error probability even if the quality of watermarked image is as poor as 26 dB in PSNR after corrupted by additive Gaussian noise.