The images are like this:
There's always 5 letters, 5 lines, A-Z, no anti-aliasing, no rotation, and Verdana 18 pt bold, and 2 pixels between each letter
I figure I should convert them to monochrome
Then simply try placing 18 pt bold Verdana characters in every pixel and record the X coordinate of every pixel where it fits entirely over black characters. I would find out if it fits over the black completely by just comparing the original image with an image that has the placed character and seeing if they changed at all. I would do this for every letter, then make sure that all of the letters had 2 pixels of space in between, and then just sort them by the x coordinate order.
Would this be an ok way to do this, or would there be a more effective way of doing it? The speed it needs to do this wouldn't really need to be that fast, jist faster than the speed of downloading the page/image, but would doing this be even slower than that?