Who knows what is going on between 12 and 11..?

[Meremail]

I had a quick browse over Wikipedias jpeg page and there is some interesting reading to be had.
At the top there is a very big image of sample images of the same shot taken at different compression settings, and a graph of file size vs compression, but note that this doesn't include (perceived) quality. The file size increases exponentially with compression reduction.

If you go down the page past enormous mathematical data, there is discussion about the effect of compression based on image content and particularly where the content has sharp edges.

So it is likely that the simple answer to Eigils question ( his image would be handy ) is in the Wikipedia graph. His comment on "what pixels are lost" is probably expressed too simplistic as the encoding/decoding process is complex and not simply throwing away information but transforming it.

Refer to http://en.wikipedia.org/wiki/JPEG

[timo2]

Tim,
This is my point. "No visible difference.." not on a normal screen at least - then what is the overflow used for - the machines printing large posters.. or maybe because we don't want to throw good image information away..? There must be a logical reason for levels 10 to 12. Otherwise my new choise will be Q10. Is there an engineer from Adobe out there

When I said no visible difference, I was meaning at 100%

FWIW I also shoot stock. Level 10 (and above) is approved by Alamy and their QC is very stringent.

[Eigil Skovgaard]

When I said no visible difference, I was meaning at 100%

FWIW I also shoot stock. Level 10 (and above) is approved by Alamy and their QC is very stringent.

And 100% was what I read.. I regard "actual pixels" as an implicit minimal precondition for technical examination of images - even though this is not a law of nature.
Alamys "10 and above" was another incitement to ask the question.

[Eigil Skovgaard]

..
If you go down the page past enormous mathematical data, there is discussion about the effect of compression based on image content and particularly where the content has sharp edges.

So it is likely that the simple answer to Eigils question ( his image would be handy ) is in the Wikipedia graph. His comment on "what pixels are lost" is probably expressed too simplistic as the encoding/decoding process is complex and not simply throwing away information but transforming it.

Refer to http://en.wikipedia.org/wiki/JPEG

Hi Jim,
Thank you for the link to Wiki - longhaired stuff! but I'll give it a try.
This should be a link to the image in question - a fairly high frequency pattern, though the beautiful details are lost with this format. It came out as a 29.83 Mb jpg at Q12 and the file size was reduced to fantastic 16.9 Mb at Q11: http://www.ask4photos.com/index/deta..._oak_tree.html

[Meremail]

Thanks for the link to the photo Elgil, I can see why the highly contrasting edges might be an issue considering what has been said in this discussion.
I got around to doing my own tests on a single image today, and have posted my results on my Flickr page at http://www.flickr.com/photos/meremail/4944893908/

I won't clog your discussion with more info here. It is an interesting exercise, and I recognise that I may have made wrong assumptions, and am open to correction.

Jim

[Eigil Skovgaard]

Thanks for the link to the photo Elgil, I can see why the highly contrasting edges might be an issue considering what has been said in this discussion.
I got around to doing my own tests on a single image today, and have posted my results on my Flickr page at http://www.flickr.com/photos/meremail/4944893908/

I won't clog your discussion with more info here. It is an interesting exercise, and I recognise that I may have made wrong assumptions, and am open to correction.

Jim

A very eyecatching opening photo on your Flickr page, Jim! Will read your test later today.
Thank you for a very helpfull contribution to this discussion.
Br, Eigil

[Eigil Skovgaard]

Hi Jim,
Had a closer look at your Protea flower -series on Flickr today. Sends my message here, as an account is necessary on Flickr.
With my 63 years eaglesharp eyes I can't see any differences between the quality steps within the available formats. Of course it would have been nice..! with a series of the same 100% crop with softly graduated surfaces and distint details as in your Protea. Got the feeling, that I had seen such a presentation before, and maybe it's this one: https://www.cambridgeincolour.com/tu...imagetypes.htm (scroll down). This is a 200% crop shown at 100, 80, 60, 30 and 10 percent quality - unfortunately these steps are not directly convertible to Photoshop's 1-12 quality steps. I must admit, that I have a hard time seeing any difference between 100 and 80 percent. On the other hand this test image doesn't show a smooth graduated surface.
Br, Eigil

[Michael Bailey]

If I'm accurately reading between the lines of some of the previous replies, it looks like a lot of this discussion is really about archiving. If so, then let me throw in this diversion:

Instead of archiving in Jpeg format and worrying about quality levels, may I suggest saving your files as Tiffs with LZW compression. LZW is lossless and your files might even be smaller than Jpegs at 11 or 12. You can do this even if your camera originals were shot as Jpegs.

If you choose someday to save in 16 bit depth, I believe the best choice is Tiff with Zip compression. These files are slow to save and open, but are best at keeping all the information in the smallest package. 16 bit will not be an issue unless you shoot in raw format.

MB

[Eigil Skovgaard]

If I'm accurately reading between the lines of some of the previous replies...

Hi Michael,
No, not directly. It was a problem that some stock sites were unable to receive jpg-files bigger than 20 Mb (though others are thrilled to be able to offer super sizes). I was reluctant to reduce the quality from 12 to 11, and even more when I got this remarkable smaller file. It troubled me, so I posted the question above. What happens to one's good images when 9 Mb of the information disappears? I still don't know the precise answer, but reducing 1 to 2 steps seems to be less harmful in the high end of the quality scale. Someone mentioned, that this reduction is invisible even at 400 percent on a normal screen, which is a fair trade I think. Maybe you have to do serious printing, before the missing 9 Mb matters.
Btw. I always work in 16 bit tiffs and save a LZW compressed copy of the edited version together with my original raw file. In a way you can say, that this is a matter of archiving - in my case while I am saving up for a better printer
Br, Eigil

[Colin Southern]

Maybe you have to do serious printing, before the missing 9 Mb matters.

I suspect not even then. Printing is pretty low dynamic range stuff and it's amazing how minor changes in adjacent levels just don't show; even quite large changes aren't readily apparent.

[Dave Humphries]

Maybe you have to do serious printing, before the missing 9 Mb matters.

The other thing to consider is that if the difference between Tif and compressed Tif loses several MB and is lossless, by the same token, that probably applies to most jpgs too - not all the 'missing 9MB' actually equates to missing picture information that can't be recreated. Not sure I explained that very well but I know what I mean

Cheers,

[Eigil Skovgaard]

I suspect not even then. Printing is pretty low dynamic range stuff and it's amazing how minor changes in adjacent levels just don't show; even quite large changes aren't readily apparent.

And gigantic banners neither I guess.
Maybe we are waiting for the ultimate screen to show the difference?
Br, Eigil

[Eigil Skovgaard]

The other thing to consider is that if the difference between Tif and compressed Tif loses several MB and is lossless, by the same token, that probably applies to most jpgs too - not all the 'missing 9MB' actually equates to missing picture information that can't be recreated. Not sure I explained that very well but I know what I mean

Cheers,

Hi Dave,
Got it the first time - uniform areas are re-stored without any loss. The loss is in averaging nearby values to be uniform. Then my image loose graduation I guess.
Br, Eigil

[Michael Bailey]

Hi Eigil,

Thanks for the clarification. Now I appreciate what you're up against. For what it's worth, I'll join those who can't find a difference between Jpeg 11 and Jpeg 12. In fact, I've never been able to tell 12 apart from 10.

I'm curious about your use of LZW with 16 bit files. When I try it, the files often become bigger than the uncompressed versions. Why? No idea.

[Eigil Skovgaard]

Hi Eigil,

...
I'm curious about your use of LZW with 16 bit files. When I try it, the files often become bigger than the uncompressed versions. Why? No idea.

Hi Michael,
I think you should make this question a new thread. I don't know the answer. Actually I am rather confused about how PS's LZW compression handle 12 versus 8 bit images.
Example: From PhaseOne I output a 16 bit TIFF file (ff), size 139 Mb. Why 16 bit? - a wise man once told me that 16 bit is best for editing. After editing in PS5 the file consists of 4 layers, and saved as a 16 bit TIFF with LZW the size is 956 Mb !! (saved with NON compression, a little smaller 930 Mb). I am refering to the file sizes shown in Bridge. If I save the same file as a PSD/16bit the size is 474 Mb. That should make PSD the obvious choise, unless you need to post in TIFF.
Well, this TIFF-filing is an old habit, could be time for a revision.
Here comes the weird part, and I am probably exposing one of my limitations here. If I, after editing, change the bit-depth of the TIFF to 8 bit and save it with LZW compression, the size is reduced to 298 Mb, which is ok compared to the un-edited size, 139 Mb.
Now the big question is: Did I loose all that I gained from the 16 bit editing?
Apparently not. If I load the LZW compressed TIFF/8bit file into PS again and change the bit-depth back to 16 and compare the image with either the PSD or the uncompressed layered TIFF version, I can find no differences. If I compare the histograms layer by layer they are identical..! - and if I compare the image quality at an enlargement where the single pixels start to show - 800% - I can see no differences either.
As a consequence I reduce the bit-depth to 8 before I LZW-compress.
Tell me where I am wrong.
Br, Eigil

[revi]

Hi Eigil,

You're right, no difference between 8 and 16 bit depths visible on screen. Until you start editing again, that is...
Normal screens are 8 bit/channel so even if you work with 16 bit/channel for editing, display is already in 8 bit, so a 16-bit version and an 8-bit version of an image will display exactly the same (within one level, due to rounding).

What you gain from 16 bit editing is less rounding errors during editing, and especially less risk of banding when stretching parts of the histogram: where 8-bit would straight-away have big holes in the display histogram, this won't happen in 16 bits, the display still can fill all (8-bit!) values.
So saving the final image in 8-bit/channel wouldn't lose anything, but then you can just as well flatten it to one layer and save as a good quality JPG... Saving intermediate edits in 8-bit to continue editing later might lose you quality, but less than repeated saving/opening in JPG would.

As for the file sizes, TIFF seems rather generous in its use of disk space: I'd expect about 4*139 or 560 Mb files for that (actually a bit less, as metadata shouldn't be duplicated per layer). As PSD is the native PS format, why not use that, and create a TIFF only if needed (and 1-layer then?)

[Eigil Skovgaard]

Hi Remco,
Thank you for a nice clarification. This really makes me consider to change from the primary TIFF out of CO5 to PSD in the first save, as there is nothing to gain with the LZW compression on 16 bit TIFFs.
Michael's question - why is LZW compression in PS not reducing but adding to the size of 16 bit TIFFs - is still open, and it could be interesting to know if anybody had a good expanation to that mystery too.
Br, Eigil

[Michael Bailey]

When I'm figuring out what to do when storing these giant multi-layer files, I often have to remind myself that what I'm really saving is not the picture, but only a record of the editing I've done to it. After all, I still have the raw file and its settings. If I shot Jpegs in camera, I imagine I'd save those original camera files separately from the edited work.

So, when I listen to the efficient part of my personality, I don't mind converting my 16 bit edited work to 8 bit before storage and save an awful lot of disk space, whether as a PSD or LZW Tiff. I think Remco is right when he says that the biggest effective difference between 8 bit and 16 bit files is what happens to them as they're edited.