Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: Edalongthepacific on January 26, 2012, 06:59:10 am
-
I processed the same image in both 8-bit mode and 16-bit mode using PS CS5. I used identical, and somewhat over-the-top, Levels settings on both images. The histogram for both the 8-bit and 16-bit processing are almost identical. In fact, the major difference in the histograms can be attributed to the difference between the "8" or "16" used in the text layers.
I have been led to believe that processing in 16-bit mode is less "destructive" but the evidence here does not support this. Both histograms show an equal amount of pixel loss due to the adjustment layer. The original image is 60 megabytes large and taken with a Canon Mark II, 21 megapixel camera.
Discussion?
-
There is no "pixel loss". The image size is the image size unless you resize it and resample it, regardless of whether the pixels are calculated at 8 bit or 16 bit depth. The differences between the two calculation approaches would tend to show on large prints of certain files through a subtle loss of quality in tonal transition especially over smooth areas of the image.
-
Given the comb histogram in 16bit mode, weren't you in 8-bit somewhere before your test? ;)
16bit processing is more useful with more than 8 bits to start with...
-
Furthermore your image data is probably already gamma-encoded (I'm guessing it's sRGB, Adobe RGB, ProPhoto RGB, or similar). Doing a Levels adjustment (even a strong one) is not likely to introduce serious problems in this context. Also, a single operation is not likely to result in problems. The problems occur when doing a sequence of adjustments (i.e., extensive editing).
-
Given the comb histogram in 16bit mode, weren't you in 8-bit somewhere before your test? ;)
16bit processing is more useful with more than 8 bits to start with...
Exactly! The OP needs to give a complete description of his workflow. For more on this topic, readers should look at Bruce Lindbloom's Dan Margulis' 16-bit Challenge (http://brucelindbloom.com/). Look under the Info tab of the main screen.
Regards,
Bill
-
Both histograms show an equal amount of pixel loss due to the adjustment layer.
And after you flatten (the same effect that would be sent to a printer)?
-
I processed the same image in both 8-bit mode and 16-bit mode using PS CS5. I used identical, and somewhat over-the-top, Levels settings on both images. The histogram for both the 8-bit and 16-bit processing are almost identical. In fact, the major difference in the histograms can be attributed to the difference between the "8" or "16" used in the text layers.
I have been led to believe that processing in 16-bit mode is less "destructive" but the evidence here does not support this. Both histograms show an equal amount of pixel loss due to the adjustment layer. The original image is 60 megabytes large and taken with a Canon Mark II, 21 megapixel camera.
Discussion?
I did a similar experiment and got different results. I selected a contrasty picture (of little artistic merit) that was taken with the Nikon D3 and rendered the image into TIFFs with ACR using a linear tone curve in 8 bit and 16 bit. I then applied a curve in Photoshop to bring up the shadows in both images (using the identical curve, which was saved as a preset). The results are shown. The manipulation was hardly extreme, but the histograms are quite different.
(http://bjanes.smugmug.com/Photography/8-vs-16-bit/i-Z59qfF4/0/O/Linear.jpg)
(http://bjanes.smugmug.com/Photography/8-vs-16-bit/i-sPBsVh8/0/O/Curve.jpg)
(http://bjanes.smugmug.com/Photography/8-vs-16-bit/i-X4X6mwb/0/O/Curve.png)
(http://bjanes.smugmug.com/Photography/8-vs-16-bit/i-Bs9Cmz2/0/O/Histo-8bit.png)
(http://bjanes.smugmug.com/Photography/8-vs-16-bit/i-fk95CGk/0/O/Histo16bit.png)
Gaps in a histogram are not the best way to evaluate an image. Whether or not the gaps would be visible depends on the nature of the image.
Regards,
Bill
-
Hi!
There is a warning sign in your histograms saying that they are not up to date. Try to klick symbols to get the histograms up to date and see if there is a difference.
Best regards
Erik
-
Hi!
There is a warning sign in your histograms saying that they are not up to date. Try to klick symbols to get the histograms up to date and see if there is a difference.
Best regards
Erik
Eric,
I presume that you are referring to the OP's histogram, which shows the cache warning. He should completely redo his experiment.
Regards,
Bill
-
I think redoing the experiment is a waste of his time and everyone elses'. This issue of 8 vs 16 bit has been discussed to death and every angle of it is pretty well known. The OP just has to do his homework on what's been said before and how to determine what suits him best, which ultimately and probably uniquely means making the right kind of test prints in the right kind of ways under the two conditions and comparing them.
-
I think redoing the experiment is a waste of his time and everyone elses'. This issue of 8 vs 16 bit has been discussed to death and every angle of it is pretty well known. The OP just has to do his homework on what's been said before and how to determine what suits him best, which ultimately and probably uniquely means making the right kind of test prints in the right kind of ways under the two conditions and comparing them.
+1. Why did anyone bother to reply?
-
As we say in French "Bonne question" - maybe just to be polite and help steer the OP in the right direction. I think that's done.
-
Bill,
You are right as usual ;-) Thanks for filling in the info.
Best regards
Erik
Eric,
I presume that you are referring to the OP's histogram, which shows the cache warning. He should completely redo his experiment.
Regards,
Bill
-
The Jr. member appreciates the wise guidance of his elders and will rethink the experiment from the ACR upward.
-
Download this 8-bit TIFF file: 8bitposterization.tif (http://www.guillermoluijk.com/misc/8bitposterization.tif), with 2 curves applied to a smooth gradient. If done in 16-bit mode the image remains unaltered, but if we stay in 8-bit mode a clear posterization will appear because of the 8-bit rounding caused by the first curve:
(http://img42.imageshack.us/img42/9720/examplev.gif)
In some cases staying at 8 bits could be OK, but in other could not, specially the more agressive your processing is.
Regards
-
Hi Guillermo,
I've tried to duplicate your sample but I'm not seeing any difference between the 8 bit and 16 bit. I only applied an "S" curve which combed the histogram, what did you use?
Gale
-
Hi Guillermo,
I've tried to duplicate your sample but I'm not seeing any difference between the 8 bit and 16 bit. I only applied an "S" curve which combed the histogram, what did you use?
My curves are extreme, and the scene is a soft gradation, very prone to suffer from visible posterization. It's a Photoshop file, just two curves that also comb the histogram.
-
Okay, "extreme", probably not something you would do to an average image then. Correct?
Gale
-
Okay, "extreme", probably not something you would do to an average image then. Correct?
Gale
Define average...
Anthony
-
I think instead of splitting hairs over definitions, it's best to accept Guillermo's astute demonstration that there are image editing conditions in which you can break an 8 bit image, while the 16-bit version will remain intact. If that's not important to you and you find 8-bit generally fine, use it; if you want the added assurance of preserving image quality under major edits and don't mind the extra storage space, work in 16-bit from start to finish. These days this is the generally recognized recommendation for image editing and it has a basis in theory and practice.
-
I redid the experiment based upon the wise counsel of the patient and courteous senior members and discovered that I was wrong. Not just a little wrong but completely wrong. There is a significant difference in 16 vs 8 bit processing. Weather this difference translates into a noticeable improvement in the final output depends upon the nature of that output and the observer.