Watermark detection in an image or the like can be optimized by exploiting the innate biases in the image to emphasize the watermark signal. The watermark signal can be trial-located with different origins in the image to find one that yields improved results. Similarly, the image can be processed (e.g., by changing resolution, rotation, or compression) so as to change the innate biases to better reinforce the watermark signal. Compression of an image can be done in accordance with a desired identifier, with the compressor deciding which image components to retain and which to discard based, in part, on the identifier that is to be associated with the image. The techniques are also applicable to other forms of content, such as audio.