Word Verification Reader, Update #1

I should be working. But I'm not. This Word Verifier thing is so much more interesting. First, I opened the url for the Word Verifier image on the comments page, and refreshed it until I found a very simple example; and my lucky letters were eiifv. I decided to do this in C to begin with. I can always switch to Python or some other language, later on. I used libjpeg to decompress the image, and then I wrote a little bit of code to separate the letters from the background in a very naive and simplistic fashion. Here's a little ASCII art that depicts the distinction the program sees.

Now that I've actually got pixels to work with, I can start thinking about edge detection, lines and letters. Fun, fun, fun.

Update: It has been pointed out to me that blogging about my progress in defeating an anti-spam measure on the blogging system I use, might be self-defeating. I'm just going to not think about that.

Update: Here's some more interesting stuff to look at. First, I blew up the eiifv image a bunch, so you can see the detail - specifically where the right side of the 'f' touches the 'v'. Then look at the output from the improved version of the program ... much crisper and more accurate. Still got a ways to go, though.

Update: It's very simplistic, and I only have the drawing code in place for vertically-oriented rectangles, but, it *is* progress. You can see in this picture that a "rectangle" of sorts was found in the top of the 'f'. The hyphens mark off all of those pixels which been "reduced", and the asterisks indicate that line to which those pixels were reduced.