The images are 2D, they do not contain 3D (Z axis) data. But imaging software is able to detect small deltas in colors, tints, and shades in the form of a color histogram. Image color histograms are a representation of the distribution of colors in an image.
Histograms appear to be 3D by assigning 'height' to the deltas in the colors. In Robert's gray image above it is tempting to think that it has detected 'height' or a embossing on the stamp but it is actually analyzing the color deltas in the stamp image. If you took a piece of multicolored steel which was perfectly flat, a color histogram would assign 'heights' to it.
The only way a watermark could be detected using this method is if there is a difference in the colors. If a stamp was perfectly monochromatic and the scanner and computer detected color deltas, it may well be a watermark. But stamps are not a single color, they are multicolor and often have cancels. If you look closely at Robert's histogram above, you can see how the cancel on the face of the stamp has been detected and is assigning 'depth' to it. (In this histogram it is assigning depth to the dark colors.) Note how the black spot appears as a pit in the histogram image.
More info on image histogramshttps://en.wikipedia.org/wiki/Image_histogram