I use image processing software (PhotoPaint, if available, and Photoshop if I must) for image data layers on a varity of occasions. sometimes image data needs correcting: Natural Earth, for example, misplaces the Balney Islands (in the high Anatarctic latitudes) an island width or two too far west. On a world map it is not noticeable, on an Antarctic map of sea ice rather more so. Cloneing the islands in the correct location and removing the misplaced ones is easy in the photo software. Because I don't change the image extent, I can easily re-georeference the image by typing the known values into the link point dialog.
working with polar projections, one gets a poor join on projecting an equator based cylindrical image (again using NE as an example). One can get rid of the ugly seam running up the 180 degree longitude quickly with PS, and pop the image back in the mapping software.
My biggest use for photo software was making forest change detection maps. I did NOT want ANY antialiasing or blurring, and very sharp edges with very exact colors on the very small blocks of pixels. For that, I would export tif images of the mapping software layout raster showing only one class of the data, usually just in black. Thus, I had a set of images of exactly the same extent, wash with one class of data. In PS, I could replace the exported color with the target color and make the background transparent. Reassembling the image layers, and flattening them, gave me a very clean and sharp base image to use in assembling the map in CorelDRAW or Illustrator. I could also intersect any hill shade, vignette, or other effect before going to the graphics program.
So, photo software is an excellent tool in map making, but it is usually a mistake to expect to use it for too much.