The visual quality of the extracted watermark is an important criterion when judging the power of watermarking scheme to protect the ownership of an image. The extracted watermark is usually damaged by the rounding operations that come in during the transformation between spatial and frequency domains. A solution for reducing the damage is not to use the rounding operations but rather some other criteria to guide each pixel value calculation through either carrying or truncation. In this article, we propose a heuristic method to enhance the quality of the extracted watermark. The image is divided into nonoverlapping blocks, and the reference coefficient data are used to modify the pixel values. Extensive experimental results show our proposed method can easily reduce the damage to the extracted watermark.