Re: OT: 16 bit editing myth or reality?
Yves, My reference below to tonality when converting a 16 bit file to 8 bit should have read "Editing a 16 bit file will leave a rich file that when converted to 8 bit mode for printing has no gaps in THE HISTOGRAM...." Instead of "Editing a 16 bit file will leave a rich file that when converted to 8 bit mode for printing has no gaps in TONE...." Since there would certainly be 16 bit tones that would have to be dropped in order to convert the file to 8 bits. Sorry if you interpreted the my reference to Ryuji's statement that working in 16 bit is a "no-brainer" to implying that you were "a brainless whatever"—it's just a figure of speech and not intended to be a reference to your intelligence—besides how would I know how intelligent you are—I've never met you. I did say that I thought that if you wished to conduct some sort of experiment in Photoshop and report the results to the list that I would be interested in what you found. In the mean time, I have Christmas shopping to do! I don't need to do any experiments to know that working in 16 bit is better, did those long ago—sorry! Ok, here is a fun one I did years ago to test Photoshop & 16 bit—and you don't even have to edit your heart out—or any other organ you value: 1. Make a 16 bit color file that is about 10 pixels high and 30,000 pixels long (Max Photoshop will do in length). Fill it with a gradient fill from 0 to 255 (black to white). 2. Duplicate the file and convert it to 8 bit. 3. Apply a Levels Adjustment to both files, bringing the OUTPUT WHITE POINT SLIDER down from 255 to 20. 4. Now for Yves sake, make the 16 bit file an 8 bit file 5. To make sure there is no cache funny stuff happening, close both files and reopen them. 6. Examine the histograms of both files (hint: the one that looks like 20 toothpicks marching is the 8 bit file that had the levels adjustment applied to it in 8 bit mode) 7. If you want, print out the two gradients, view them from across the room and see if you can tell one from the other... the one that looks like a 21 step tablet is the 8 bit file that had the levels adjustment applied in 8 bit mode. Happy Holidays All ! Mark Nelson Precision Digital Negatives - The System PDNPrint Forum at Yahoo Groups www.MarkINelsonPhoto.com In a message dated 12/18/07 5:01:30 PM, gauvreau-yves@cgocable.ca writes: Mark, I'm sorry to say this but the second phrase you wrote below is not true in the strict sence of words. In fact it's practically sure there will still be gaps in 16 bit edited file, I'll admit they could and probably will be much less then it would have if all editing had been done in 8 bit, I have no problem with that, my first message is the proof of that. Comparing me to a brainless whatever doesn't proove that I'm wrong. Have you tried the little experiment I suggested, it will take you less then 10 minutes? In case it's already in the waste basket here it is again. Find yourself a hi bit color image, one your sure it as more then 8 bit depth (as you said yourself below, some claims of hi bit depth can be deciving). Make a copy of it and convert the copy to 8 bit, to make sure PS doesn't just put a tag on the image and keep the hi bit values in memory, save the copy, close it and reload it. Now you are sure you have both an 8 bit and a hi bit image to work with. For the effect to be more convincing I suggest you don't use any form of random editing or one you can't repeat exactly on both image, in principle a random process (gaussian stuff) can not be repeated twice and for the effect to be more evident I recommend not to use them for now. Now you can edit your heart out on both image as long as you do the "exact" same thing on both image I also suggest you be reasonable, do as you would in a normal situation, it would be easy to do things to fool the test but you wouldn't normaly do this in your usual workflow, when your done convert the hi bit image to 8 bit and just to make sure you truly have an 8 bit image, save it, close it and reload it. Next, click on image->apply image... choose substract and set the offset to 128, click ok (don't forget to select the other image). Now at what magnification do you begin to see some difference in levels in your test image? (I wasn't able to see anything until I was at 200% mag with carefully chosen edits and at 100% with some carelessness and I'm an old guy my vision is not that of a teenager at its prime) You can also look at the histogram which will give you some data you could use to evaluate the test. If you applied the exact same editing steps to both images and no random steps, the histogram will tell you something like this, the vast majority of the millions of pixels in the resulting image are at or very near 128 (A - B = 0 + 128 = 128) and within 3 to 5 levels (more or less depending on how carefull you where) on each side there will probably be nothing to see. There are some small different between the images which implies there is probably a difference in both type of editing, the histogram prooves that but where the edits the exact same? Now these relatively small difference are spread all over the area of the test image and if you where to view the two final image side by side you probably couldn't see any of these difference on your monitor unless of course you know where to look for or if you magnify to the sky. Now lets raise the bet sort of speak, lets print those 2 images at whatever size you like, I ask you only one thing, view them at the recommended distance based on their size, fair enough, ask anybody around you to compare the two image and ask them to tell you which one was edited in hi bit mode or which one looks better if you which, if the results of doing this test with a couple dozen of people differ significantly from using a coin flip to decide which one was edited in hi bit, I'll buy your book, just to say it's not a bet. Lets come back to the real world for a minute, above I asked you not to use random editing because you can't do the same exact edit twice. In the real world it would be almost impossible to edit two image in the exact same fashion (which probably happened already above), we have these random edits and we also have local edits and whatever else you can think of that are quasi impossible to do twice exactly the same way. In the test above I tried to elimimate these as much as possible to reduce the difference to its minimum. In real life it would be surprising someone would loose his time to do this but lets say one those (any normal edits), the only difference would be a wider histogram and then you could say Ha! Ha! I told you so. It happen I did it, yes, so I know and the difference are still very hard to see at 200% magnification on the actual images on the screen, I wasn't able to see them but don't take my word for it, try it yourself (which is probably what you did already). By the way, have you or anyone else for that matter, thought that converting a 16 (15) bit image to an 8 bit one mean a compression of at least a 100 to 1, it's impossible not to loose quite a bit in the process, just remember that. Happy holidays Yves PS For the others on this list, I wont buy your stuff but I defy you to "proove" what you say with facts (not just words) and thus proove me wrong at the same time. ----- Original Message ----- From: Ender100@aol.com To: alt-photo-process-l@usask.ca Sent: Tuesday, December 18, 2007 1:31 PM Subject: Re: OT: 16 bit editing myth or reality? Yves, It is true that most printers currently print in 8 bit mode and even if you send a 16 bit file to them, the file is converted to 8 bit mode—HOWEVER... Editing a 16 bit file will leave a rich file that when converted to 8 bit mode for printing has no gaps in tone.... Editing in 8 bit mode will leave gaps that will still be there when printed, possibly causing posterization. As Ryuji stated, this issue is really a "no-brainer" and has been considered such for years by the experts. I'm not sure what your point is other than maybe setting up a model to illustrate what the actual final difference is in tonal richness for the same file in both 8 bit & 16 bit mode when given the same adjustment. If you can come up with a model that would demonstrate it, it would be interesting to know, but it won't change the reality that 16 bit files are superior. By the way, supposedly the newest Epson printers working with Mac Leopard will be able to make "16 bit" prints as soon as Epson releases the drivers to do so. Also one of the other printer manufacturers has a plugin that is supposed to give around 14 bit prints......so the 8 bits/printer thing may soon be a thing of the past. Another side note, a custom scanning house of "good reputation" was making drum scans for people that were supposedly 16 bit drum scans.... in fact, they were scanning in 8 bit because that was all their drum scanner would do, then converting them to 16 bit and sending them out. This is, however, rather easy to detect. Of course they were charging extra for their "16 bit scans". Happy Holidays, Mark Nelson Precision Digital Negatives - The System PDNPrint Forum at Yahoo Groups www.MarkINelsonPhoto.com In a message dated 12/18/07 12:18:13 PM, gauvreau-yves@cgocable.ca writes: All this above is true while in hi bit mode but the question is what happen to these extra bits when you need to print the image knowing the majority of consumer printers out there are 8 bit printers? The simple answer is, they go back to the black hole they came from. The true benefit of editing in hi bit mode is so small, it is for all practical purpose insignificant on actual prints but I'll give you this it looks much better while in this hi bit mode. ************************************** See AOL's top rated recipes (http://food.aol.com/top-rated-recipes?NCID=aoltop00030000000004) = ************************************** See AOL's top rated recipes (http://food.aol.com/top-rated-recipes?NCID=aoltop00030000000004)
|