Image -> Ascii conversion algorithm idea

I was thinking (honest! .. it DOES happen *g*) when I saw JavaJulia by RMG (the
Geiss-guy) that the same algorith he uses to convert julia-fractals to ascii
could also be used for any graphic.

JavaJulia uses the chars " ,._'/)J`(\L"7PO" to creates neat Julia fractals.
(Btw he could have used � and many other extended chars since the medium was a
webpage)

I have been studing the code (because I thought it was intriguing and I wanted
to learn how he did it) and it seems that for each char it determines if 4
points are a part of the fractal (white) or if they are not.
Depending on which points are white and which are not it represents theese
points with a certain char from the chars mentioned earlier.

Some examples :
   __
  |##|  = 7
  | #|

   __
  |  |  = _
  |##|

   __
  |##|  = O
  |##|

Knowing this we can create a similar effect to any black & white image.
This works great for illustration and other graphics in only 2 colors.

But what is we wanted to apply this to a grayscale image (256 shades of grey) ?

I had an idea that if we instead assigned a value (0 - 255) to theese points
instead of just 0-1 (black or white) we could apply the above technique
(choosing the right char) to thoose points (pixels) IF and only if the
difference in brightness (0 - 255) was above a certain level (user adjustable)
and if not then the points brightness value would instead be averaged and a
"soft" char would be choosen to shade the area.
With soft char i mean a char without a distinctive shape .. like f.x :  #8Oo:.
( possibly userdefinable too ) ..

It's all very much like a mix between Jave's 4-pixel and 1-pixel algorithm.

I believe that this algorithm will yield very good results on illustrations and
high-contrast/low-detail images.

RFC

Sincerly
CeeJay