In just a few years, robustly embedding an imperceptible watermark into a digital image for the purpose of ownership verification and copy control, has been widely investigated and studied. One of the remaining challenges for robust image watermarking is to design secure encoding/decoding schemes that reliably verify the embedded but possibly corrupted watermark.In this paper, we introduce a modified embedding scheme that can render more reliable watermark detection compared to some conventional image watermarking approaches. We first study common problems inherent in traditional watermarking, and then propose a better embedding rule as the remedy. Experimental results demonstrate that the new embedding rule enhances the performance of existing watermarking scheme for different capacities/payloads and under various types of common image processing.