A novel framework is presented for image watermarking, which enables detection of the watermark without the original image. As opposed to earlier DCT-based subsampling method that reconstructed the image before transmission, the new DWT-based framework allows validation of the watermark before the recovery of the image. This reduces computational requirements in situation when the verification step fails. And compared with previous subsampling approach, this algorithm apparently preserves superior robustness under various attacks, especially under JPEG and lowpass filter, which is the relative weakness of the prior method.