In most watermarking schemes for copyright protection, a seller usually embeds a watermark in multimedia content to identify a buyer. When an unauthorized copy is found by the seller, the traitor's identity can be traced by the embedded watermark. However, it incurs both repudiation issue and framing issue. To solve these problems, some buyer-seller watermarking protocols have been proposed based on watermarking scheme in the encrypted domain. In this paper, an enhanced watermarking scheme is presented. Compared with Solanki et al.'s scheme, the enhanced scheme increases effective watermarking capacity, avoids additional overhead and overcomes an inherent flaw that watermarking capacity depends on the probability distribution of input watermark sequence. Based on the security requirements of buyer-seller watermarking protocols, a new watermarking scheme in the encrypted domain with flexible watermarking capacity is proposed. It improves the robustness of watermark sequence against image compressions and enables image tampering detection. Watermark extraction is blind, which employs the same threshold criterion and secret keys as watermark embedding. Experimental results demonstrate that the enhanced watermarking scheme eliminates the drawbacks of Solanki et al.'s scheme and that the proposed watermarking scheme in the encrypted domain outperforms Kuribayashi and Tanaka's scheme. © 2010 Elsevier Inc. All rights reserved.