LSB steganography is trivially easy to detect. Do not use LSB steganography.
Steganography can be provably secure, with the correct amounts of plain text and "cover data".
Here's one PDF. (http://www.cs.cmu.edu/~biglou/PSS.pdf)
Some people need this kind of stuff to work to avoid torture or other serious consequences, but most implementations are toys to demonstrate proof of concept or to play with the ideas.
tl:dr a real cryptographer will be along to correct my mistakes and provide sensible links.
That said: the PDF you link to seems to show that provably secure steganography (as defined by this paper) exists if, and only if, one-way functions exist. (See section 5.3)
It so happens that if one-way functions exist then, as a corollary, P != NP. In other words: if steganography is provably secure, P != NP.
So let's not get ahead of ourselves. ;-)
Never learned the back stories of other famous images that may be older than "Lenna", e.g. "Girl" and "Cameraman".
Quick improvement to the LSB method: Use JPEG and embed the data in the LSB of the DCT coefficients (still detectable, but not as trivially)
Cool demo though - I'd add a disclaimer though. It's easy to think that it's obvious that people shouldn't use this to store their password in, say, their Facebook profile photo, but..
EDIT: Though if you were really secure, you shouldn't be typing your passwords into a third party website in plaintext. Alas, I'll have to write my own.
A big QR-code in your table is something obvious to scan, and aesthetically ugly. A framed picture not so much, and could serve a similar purpose.