[Digikam-users] re JPEG lossiness, PNG

[Remco Viëtor]

On Saturday 14 January 2012 20:26:53 Jean-François Rabasse wrote:
> On Sat, 14 Jan 2012, Thorsten Schnebeck wrote:
> >> If one selects the best, 100% quality, there won't be data loss but
> >> the
> >> final file size won't be that small, comparable to the size of the
> >> same
> >> image saved in PNG format.
> > 
> > Nope, 100% jpeg quality still creates a minor loss of information.
> > http://www.jpeg.org/faq.phtml?action=show_answer&question_id=q404fa5e29a
> > eb6
> > 
> > Bye
> > 
> >  Thorsten
> 
> Hi Thorsten,
> 
> Hem, yes and no. It's not inherent to the standard but implementation
> dependent. Since 1993, and during the 2000s, many enhancements have been
> proposed and developed for the compression scheme.
> 
> See, e.g. http://www.jpeg.pro/articles/132599/Lossless-Further-Compression
> 
> But only some JPEG generation software implement them because in most cases
> they are useless. As the article you mention states clearly, concerning the
> final result, "certainly as far as the human eye can detect".
> 
> Probably the ambiguity comes from the definition of what is lossy or
> lossless. Or, what exactly is lost ?

Afaik, there's no ambiguity in the case of compression algorithms: lossless 
means that the binary data recovered after a compression-decompression cycle 
is exactly the same as the original data.

> I agree the word "data" is not accurate and often used as a "joker" word.
> Do we speak about "binary data" loss or "quality" loss ?
> Binary data is expected to be somewhat modified, because all heuristic
> schemes have different solutions to represent the same image.
> But the important thing for us, photographers, is the visual image quality,
> colours rendering, brightness gradients, etc.
> And this can be achieved without modifications and with the guarantee that
> your image will stay as you have it, as you shot it.
> 
> But ok for my lexical unaccuracy, I'd replace data by quality.
> 

And the problem is exactly there, when images are edited: one jpeg 
compression/decompression cycle might not induce _visible_ artifacts, but it 
seems we agree it can cause 'binary' artifacts. Those artifacts will increase 
with each compression cycle (edit session), leading finally to visible 
degradation of the image (and personally, I really dislike the kind of 
artifacts it causes).

I prefer using png for intermediate edits (with the space penalty that 
implies), also so i can easily see which image to edit (raw, png: editable, 
jpeg: final). Thus no need to mess with filenames etc. for non-final images 
(but that's not a quality issue).

Regards, 

Remco