Developing a custom map means generating a potentially thousands of scaled and tiled images. convert almost makes this too easy to do.
Here's a few recipes I've used so far:
# Convert a PDF to a PNG at a resolution given by 'density' convert.exe -density 7 base.pdf base_007.png #  convert.exe -density 100 base.pdf base_100.png #  convert.exe -density 800 base.pdf base_800.png # 
In each of the above cases, I'm generating a .png from the high quality PDF file. By varying the density, I can control the resulting PNG file. In my case,  generates a 256 pixel wide image, while  generates a 28,800 pixel image.
# Slice an image into 256x256 tiles convert -crop 256x256 map.png tiles/map.png
The above command may not look like much, but the result is a tiles directory filled with 256x256 images. The file name is map-N.png, which I'll no doubt write a script to convert to map-Tx-Ty.png, where Tx and Ty are the appropriate tile coordinates.
# Convert a rectangular image to a square one convert.exe map.png -background "#CCCCCC" \ -resize 256x256 -background "#CCCCCC" \ -compose Copy -gravity center -extent 256x256 \ map_square.png
In the Google Maps examples, the base map used is square. To make playing with those examples, I use the above convert command to pad my images with a gray background. The above incantation was inspired by this question.
Oh, and it's probably worth mentioning that all these commands work just fine at a cygwin prompt, which is where I've been executing them.
If you want to slice and dice images, you need to learn ImageMagick. Period.