But why does the onion look like it's being pinched?

Humans don't really see objects - we see edges and shadows, and interpolate the objects from the shapes the edges create. One of the ways that the cells in the retina are specialized to see edges is by neighbor inhibition - if one cell is strongly responding to a light stimulus, it will send out signals that decrease the response of the cells around it. So (sorta like in Conway's Game of Life), light-stimulated cells are "louder" if they aren't surrounded by other light-stimulated cells. In the real world, this helps us see shadows cast by one object in front of another, and helps us see the shading difference that happens at a corner or a curve. (Check out some inhibition-based optical illusions).

Artists take advantage of this to create a three-dimensional picture in a flat medium. Drawn objects are shaded more darkly or more brightly around an edge-transition, and the visual system of the viewer interprets this as a non-surrounded stimulus (namely, that there's nothing immediately next to that area on one side, or that that area is presenting a different facet to the light). (Check out some shaded cubes).

So far so good. However, scanners have their own algorithms for finding and parsing edges, and, since many scanners are optimized for OCR (that's optical character recognition, not online community representative), they tend to kick up the contrast, making light areas lighter and dark areas darker. What was a gentle shade difference to enhance the 3-D-ness of the object becomes a pinched onion.

I could redraw it, or just document the bug.

(The onions look flat, of course, because they are cipollini onions. The secret ingredient in my sausage and peppers al polenta).