My idea is to grab each pixel, analyze the 255,255,255 value and give each pixel a chance to be in only 1 of 10 division I will have laid out.
This won't bring a full color representation, but my point is to make a ASCII that at least resemble the shapes of the objects in the pictures. Outline it so to speak.
Would this work?
It would work, here's a complete example in c#
http://www.codeproject.com/KB/web-image/AsciiArt.aspx
Most implementations of don't convert one pixel to one character. Instead they divide the image into rectangular regions of pixels, and then analyse each region. As well as the overall intensity of the region, you also want to look at how the intensity is spread over the region and pick a matching character.
If you limit your character set to just 10 characters however, you probably won't get a particularly good representation of the original image.
If I've understood the process correctly the effect would similar to the "posterize" option you get in some paint packages where the number of colours is reduced to some arbitrary number.