This kind of steganography-- hiding data in the least significant bits of images-- is _very_ easily detected by statistical methods, and there are many papers and tools (stegdetect, for jpeg as an example) to do so.
At a minimum, something hoping to perform successful image steg embedding should be using wet paper codes.
For one thing, I'm really happy to get your input about this implementation! Perhaps I'll try out the wet paper codes within the first packaged released of this project.