Digital image watermarking is a method of embedding a secret, imperceptible signal - a watermark - into the original image data to facilitate copy control, distribution tracking, etc. Robust watermarking schemes enable reliable verification of an imperceptible watermark, even after the watermarked content has undergone various forms of unintentional or malicious modification. The disadvantage of making a system more robust is an increase in the perceptibility of the embedded watermark. This can be overcome by using visual masking schemes that more effectively hide the watermark from the human observer. This paper presents a method that reduces the perceptibility of a robust watermarking scheme, viz. equalized-spectrum watermarking, using human visual models. The visual mask is generated using Watson's model, which is based on the block discrete cosine transform (DCT). The generated mask is applied to the watermarked image as a post-watermarking process. An existing masking equation, which is commonly used for spatial masks, was modified such that it could be used for the block-based DCT mask.