Re: filmscanners: File format

[Arthur Entlich <artistic@ampsc.com>]

I have to agree with Hugo here.  His explanation is what I always read 
and learned, and it also makes perfect sense.

The only very slight disagreement I'd have is that it might be possible 
that if you open and save in JPEG format enough times, you might 
actually reach a point where you are "chasing your tail" and that 
certain random pixels will change one way one time and then return to 
the same position the next.  However, by the time the process reached 
this point, it would seem you'd have such an abstraction of the 
original, it wouldn't much matter ;-)

The JPEG compression algorithm isn't intelligent enough to know the 
image you are currently trying to save has been previously JPEG 
compressed, so it "shouldn't do it again".  It sees an image as a 
starting point for compression, and it does everything it can to 
simplify it further in terms of the amount of data, in terms of what the 
algorithm was designed to "lose" as non-essential information.  Since 
each time the file is "simplified" (certain detail is eliminated or 
"averaged"), the next time the algorithm "looks" at the file, it 
"assumes" that the file it is working with can be further simplified 
without major loss, because the "assumption" is the file is in a 
visually useful format when it arrives as input to the JPEG algorithm. 
It doesn't "know" the history of the file (that it might have been 
previously jpeg compressed a dozen times).

Lossey compression schemes are the digital equivalent to generational 
loss in analog systems, except they use more intelligent methods to 
"lose" data by taking away that which is considered least 
noticable/significant by human perception first, while analog is much 
more random.  Also while analog systems both lose and gain information 
(noise) during generational loss, digital lossey compression systems 
mainly subtract information, to simplify and reduce the amount of data.

Art


> You lose information every time - just as Henk de Jong's attached images
> showed. You can all try it. Just try it with a heavy compression, then
> you'll be able to see the effect easier. Of course, the same applies to
> the less heavy compression, in which case you could do this a couple of
> times and not notice any degration of the image. (Yes, we might over do
> all this, but his is how it works...)
> 
> It doesn't matter if you changed anything in the image or not. Well,
> generally. You might have some programs that let you chage the heared
> portions of the image file. If you do not change any value of any pixels
> in the image, why should you then save the image? Just close it. Then you
> will not lose anything. Or, if you use lossless compression (like tiff),
> then your recreation of the image is perfect, and that's why you don't
> loose anything with opening the saved image.
> 
> It's a different thing if you are working on an image in PS and saving it
> between every step. As long as you don't open the image (or do revert?)
> you will not lose information in this way.
> 
> Best regards,
>       Hugo.
>