Luminous Landscape Forum

Raw & Post Processing, Printing => Digital Image Processing => Topic started by: DeeJay on January 05, 2011, 09:01:02 pm

Title: 8 or 16 bit
Post by: DeeJay on January 05, 2011, 09:01:02 pm
Hi there,

I've always converted RAWs to TIFs in 16 bit, Even though the files are only 8 bit to begin with. I was told that this is the best thing to do for retouching. Can anyone please confirm this?

Thanks,

DJ
Title: Re: 8 or 16 bit
Post by: Peter_DL on January 05, 2011, 09:20:04 pm
Quote from: DeeJay on January 05, 2011, 09:01:02 pm
I've always converted RAWs to TIFs in 16 bit, Even though the files are only 8 bit to begin with. I was told that this is the best thing to do for retouching. Can anyone please confirm this?

You have 8 bit RAWs ?

Peter

--
Title: Re: 8 or 16 bit
Post by: D_Clear on January 05, 2011, 11:02:37 pm
Hi Deejay,

RAW files are typically 12, 14, or more bits. If yours are 8 bits you may want to review your workflow something is amiss.

At the risk of inciting the small but vocal 'eloquent-yet-misinformed' group in this forum, the short answer is yes, generally speaking it is better to begin retouching/finishing with the maximum information in the file, which would be the 16 bit option in PS.

In crude terms, the path from capture to print, etc., is a reductive one from the standpoint of image data, it makes sense to discard information only by choice, or when necessary. Given this, starting with the most information possible for any given file is a good strategy.

Though the higher bit rate is often considered to be unseen due to some monitor, or printer, or visual judgment limitations, one of the practical reasons why we work with files in this fashion is because it will handle more manipulation in PS without suffering deterioration as can be seen at lower bit rates.

In my work we see this all the time when we shoot D3X and Phase P65+ side by side on assignments, the D3X as great as it is, will start to show 'cracks' while the Phase will not.

Hope this helps you

DC
www.dermotcleary.com
Title: Re: 8 or 16 bit
Post by: DeeJay on January 06, 2011, 01:24:23 pm
Thanks, yes that was the reason I had believed also.

And sorry, correct, they are 12 bit files. I was confusing myself with the choice between 8 bit and 16 bit conversions.

Thanks,
Title: Re: 8 or 16 bit
Post by: Peter_DL on January 07, 2011, 08:07:40 am
Quote from: DeeJay on January 06, 2011, 01:24:23 pm
... correct, they are 12 bit files. I was confusing myself with the choice between 8 bit and 16 bit conversions.

Yes, Raw files are typically 'high' bit (> 8 ),
and subsequent 'heavy lifting' in the Raw converter is supposed to be high bit as well,
independent whether 8 or 16 bit are selected for output (of course we can't know for every software around).

Then it all depends on the purpose:
rendering from ProPhoto RGB to print is certainly better done at 16 bit,
whereas 8 bit may be enough for some retouching on sRGB files for a slide show or a Frontier print.

Peter

--
Title: Re: 8 or 16 bit
Post by: Guillermo Luijk on January 07, 2011, 04:05:01 pm
Just some comments:


Regards
Title: Re: 8 or 16 bit
Post by: sniper on January 08, 2011, 06:10:56 am
Quote from: Guillermo Luijk on January 07, 2011, 04:05:01 pm
Just some comments:

  • Not all RAW files are >=12 bits. Compact cameras usually have 10-bit RAW files (enough for them since noise dithers risk of any posterization). The Leica M8 has 8-bit non-linearly encoded RAW files (again enough for them).
  • PS is NOT a 16-bit editing tool, it's 15-bit. As soon as we open a 16-bit file in Photoshop half the levels are lost by rounding to 15 bit (32768 possible values). This is still a robust bitdepth for edition.
  • Even if a source image is 8-bit (e.g. a JPEG file), it's recommended to switch to 16-bit at the very beginning if heavy processing is to be applied to it, specially when converting to wide gamut profiles (e.g. Prophoto RGB).

Regards

I am led to believe this only puts a 16 bit "wrapper" on the 8 bit jpeg and doesn't actually make it a real 16 bit?

Wayne
Title: Re: 8 or 16 bit
Post by: Schewe on January 08, 2011, 12:32:33 pm
Quote from: Guillermo Luijk on January 07, 2011, 04:05:01 pm
  • PS is NOT a 16-bit editing tool, it's 15-bit. As soon as we open a 16-bit file in Photoshop half the levels are lost by rounding to 15 bit (32768 possible values). This is still a robust bitdepth for edition.

Actually, PS is 15 bit plus 1 level...the range of values is 0-32768 for a total of 32769 levels...the reason being (as stated by some PS engineers) is it's helpful to have an exact middle value for processing algorithms...
Title: Re: 8 or 16 bit
Post by: Bart_van_der_Wolf on January 08, 2011, 12:42:30 pm
Quote from: Schewe on January 08, 2011, 12:32:33 pm
Actually, PS is 15 bit plus 1 level...the range of values is 0-32768 for a total of 32769 levels...the reason being (as stated by some PS engineers) is it's helpful to have an exact middle value for processing algorithms...

There are probably other reasons as well (because 2^16-1=65535 would have the same benefit), e.g. accomodating for some overflow without loss of precision during calculations.

Cheers,
Bart
Title: Re: 8 or 16 bit
Post by: ejmartin on January 08, 2011, 12:54:26 pm
Working with difference images would be a real pain if you didn't allow negative values.
Title: Re: 8 or 16 bit
Post by: Guillermo Luijk on January 08, 2011, 04:24:45 pm
Quote from: sniper on January 08, 2011, 06:10:56 am
I am led to believe this only puts a 16 bit "wrapper" on the 8 bit jpeg and doesn't actually make it a real 16 bit?
The point of switching to 16 bit is not making it real 16 bits, but making it more robust against strong pp.

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)

Switching to more than the original 8 bits will not improve the effective bitdepth of any resulting image, but will prevent posterization as in the extreme example above.

Regards
Title: Re: 8 or 16 bit
Post by: Peter_DL on January 09, 2011, 10:40:48 am

Great demonstration !
The curve settings are quite strong, however, it illustrates the point very well.

Thanks Guillermo.

Peter

--
Title: Re: 8 or 16 bit
Post by: geodome99 on January 15, 2011, 02:26:28 am
This is one of the most widely discussed topics at seminars and industry conferences.

I had recently read a very informative article by Dan Margulis, who by the way is one of the industry's leading pioneers and was among the first individuals inducted into the Photoshop Hall of Fame.

The article was first published in September 1997 in Computer Artist Magazine and appeared in Professional Photoshop Fourth Edition in 2002 as well as the Makeready Archive, Column 25.  The article is entitled 'Resolving the Resolution Issue'.  I suggest that everyone read it.

Quoting a small portion of the article:

'As for the gradient, just because an orange
tastes good doesn’t mean that a keyboard
does. A gradient is original, first-generation
art, from which any variation is an error.
A digital photograph is anything but; it’s
already been mangled by whatever device
captured the data, not to mention whatever
screwup the user has added.'

Hope this helps clarify things a bit.
Title: Re: 8 or 16 bit
Post by: digitaldog on January 16, 2011, 01:21:21 pm
Quote from: geodome99 on January 15, 2011, 02:26:28 am
This is one of the most widely discussed topics at seminars and industry conferences.

I had recently read a very informative article by Dan Margulis, who by the way is one of the industry's leading pioneers and was among the first individuals inducted into the Photoshop Hall of Fame.

The article was first published in September 1997 in Computer Artist Magazine and appeared in Professional Photoshop Fourth Edition in 2002 as well as the Makeready Archive, Column 25.  The article is entitled 'Resolving the Resolution Issue'.  I suggest that everyone read it.

Quoting a small portion of the article:

'As for the gradient, just because an orange
tastes good doesn’t mean that a keyboard
does. A gradient is original, first-generation
art, from which any variation is an error.
A digital photograph is anything but; it’s
already been mangled by whatever device
captured the data, not to mention whatever
screwup the user has added.'

Hope this helps clarify things a bit.

Dan’s wrong! And as another Hall of Fame member (along with Jeff Schewe who will agree with me), you can dismiss this idea of Dan’s as discussed so well here regarding his high bit challenge:
http://www.brucelindbloom.com/index.html?DanMargulis.html

So all the camera manufacturers, the scanner manufacturer’s, those that support high bit editing (Photoshop to Lightroom to Capture 1 to Bibble etc), those that provide 16-bit output to the print Path (Canon, Epson) are wrong? This is a long standing Dan belief system that only Dan and those who read and accept what he says without looking farther believe, a belief system more about getting attention than about proper digital imaging concepts.
Title: Re: 8 or 16 bit
Post by: Eric Myrvaagnes on January 16, 2011, 02:39:52 pm
Oh, goody! It's been a while since Dan M has been mentioned here. He's always good for a little entertainment.

Eric
Title: Re: 8 or 16 bit
Post by: digitaldog on January 16, 2011, 03:00:31 pm
Quote from: Eric Myrvaagnes on January 16, 2011, 02:39:52 pm
Oh, goody! It's been a while since Dan M has been mentioned here. He's always good for a little entertainment.

Yup, some urban legends refuse to die <g>
Title: Re: 8 or 16 bit
Post by: geodome99 on January 16, 2011, 03:04:37 pm
Pissing matches are unprofessional.

I still don't see any proof that 16 bit is that much better.
Title: Re: 8 or 16 bit
Post by: digitaldog on January 16, 2011, 03:14:58 pm
Quote from: geodome99 on January 16, 2011, 03:04:37 pm
I still don't see any proof that 16 bit is that much better.

Much better or any better?

The problem with simplistic statements about this is:
1. We have no idea where the data will be output. Dan shows an “example” in his halftone printed book that proposes to prove there is no difference to the data loss converting to and from Lab 22 times (as if the number of times one converts is key here, its not, the data loss occurs the first time in greatest number). What what about output to a fine art ink jet? An Ink jet or other output device with 2009, 2011 or 2015 technology?
2. What if the image is edited any number of times after? Do we know how well 24 bits holds up on one or 100 differing edits?
3. What about subsequent color space conversions (we have to do this at least once to print the damn thing).
4. What about the simple math that provides more bits than we can use when we do all of the above in high bit but not in 24 bit?
5. The fact that most every modern capture device provides more than 24 bit color anyway? So we should not use that data provided? Why are these devices high bit?

FWIW, years ago I did post a real image to his list that illustrated visible data loss with a simple edit, but because it was in ProPhoto RGB, the rules of the challenge changed once again and was dismissed. The Lindbloom piece illustrates that kind of tactic.
Title: Re: 8 or 16 bit
Post by: geodome99 on January 16, 2011, 03:26:19 pm
I'm from Missouri.

Show me.
Title: Re: 8 or 16 bit
Post by: digitaldog on January 16, 2011, 03:34:52 pm
Quote from: geodome99 on January 16, 2011, 03:26:19 pm
I'm from Missouri.

Show me.

View the example images I provided to Dan’s list years ago. They are on my idisk. 16bit challenge folder. You can even load the DNG and run the tests yourself. Render 8-bit and 16-bit, run the action to sharpen the image.

My public iDisk:

thedigitaldog

Name (lower case) public
Password (lower case) public

Public folder Password is “public” (note the first letter is NOT capitalized).

To go there via a web browser, use this URL:

http://idisk.mac.com/thedigitaldog-Public
Title: Re: 8 or 16 bit
Post by: Schewe on January 16, 2011, 03:40:59 pm
Quote from: digitaldog on January 16, 2011, 03:14:58 pm
FWIW, years ago I did post a real image to his list that illustrated visible data loss with a simple edit, but because it was in ProPhoto RGB, the rules of the challenge changed once again and was dismissed. The Lindbloom piece illustrates that kind of tactic.

See, that's the problem with the way Dan set up the argument that 16 bit doesn't offer any benefit over 8 bit.

His "rules" were, well, tuned to make 8 bit processing just about the same as 16 bit. You couldn't add a gradient based adjustment because the gradation was synthetic. You couldn't use a super-wide color space because, well, Dan doesn't believe ProPhoto RGB is a color space a professional should use. BTW, he also has stated that he doesn't think Camera Raw is a professional tool because a lack of a pure luminance based curve, so he suggests professionals process images at flat defaults and do all your tone and color correction in Photoshop (where he makes his living).

It's pretty easy to prove that using ProPhoto RGB in 8 bits/channel is suboptimal. Adobe RGB and sRGB are less bad when you start with a decent 8 bit image but you can break an image and introduce banding in both color spaces in 8 bit. You can also set up a scanner (Dan doesn't do a lot of digital capture–he's not a photographer you know). If you do base tone and color corrections in a scanner and then do light edits in sRGB or Adobe RGB on the 8 bit/channel scans, you'll prolly be ok...

Course, my preference is to use ProPhoto RGB, in 16 bit/channel, do most of my optimizing in Camera Raw (or Lightroom) and work in Photoshop in ProPhoto RGB and maintain the full 16 bit for the final images. I also print out of Photoshop using 16 bit output to my Epson printers on Mac so I never need to drop to 8 bit unless I'm preparing images for the web. I don't have banding or posterization issues in my images...YMMV.
Title: Re: 8 or 16 bit
Post by: digitaldog on January 16, 2011, 03:51:45 pm
Quote from: Schewe on January 16, 2011, 03:40:59 pm
See, that's the problem with the way Dan set up the argument that 16 bit doesn't offer any benefit over 8 bit.

Bruce summed it up perfectly:
Quote
If one takes this technique to its logical conclusion, Dan's 16-bit challenge would become "When considering all images showing no 16-bit advantage, 16-bit images show no advantage."
Title: Re: 8 or 16 bit
Post by: alain on January 16, 2011, 03:52:00 pm
Quote from: geodome99 on January 15, 2011, 02:26:28 am
This is one of the most widely discussed topics at seminars and industry conferences.
...

Probably after lots off beer.

The cost of 16-bit is so low it's a non issue. It would be foolish not to use.


I do remember 9-pin dot-matrix printers and yes it was readable,  quality wise I do prefer 600ppi laser printers.
Title: Re: 8 or 16 bit
Post by: geodome99 on January 16, 2011, 03:53:26 pm
Hi Andrew

Thanks for the link.  I will most definitely have a look.

I'm not really a Dan M advocate, or anybody else's for that matter.  I am just trying to find out the differences between the 2 ideals.

Cheers

G
Title: Re: 8 or 16 bit
Post by: Alan Klein on January 17, 2011, 12:46:19 am
Slightly off-topic, but is there any advantage to scanning film with a flat bed scanner using 16 bit rather than 8 bit?  Or is the scan process inaccurate enough so 8 will do?  (All the scans I did to date are 8 bit but you guys are getting me thinking I should change to 16 bit in the future.).  What will happen to my file size?  (at 8 bit, 2400 resolution,  120 film scans around 100meg.)
Title: Re: 8 or 16 bit
Post by: digitaldog on January 17, 2011, 12:57:43 pm
Quote from: Alan Klein on January 17, 2011, 12:46:19 am
Slightly off-topic, but is there any advantage to scanning film with a flat bed scanner using 16 bit rather than 8 bit?  Or is the scan process inaccurate enough so 8 will do?  (All the scans I did to date are 8 bit but you guys are getting me thinking I should change to 16 bit in the future.).  What will happen to my file size?  (at 8 bit, 2400 resolution,  120 film scans around 100meg.)

IF you were to do NO additional work on the image, if you got the scanner to scan in all corrections (which we assume is happening high bit) AND converting to the output space, there would be no advantages. But that’s a very, very unlikely scenario. Its probably the case you’ll do corrections and convert the data to multiple differing output color spaces. So you’d want to scan and save out high bit.
Title: Re: 8 or 16 bit
Post by: Alan Klein on January 17, 2011, 01:01:16 pm
Would the 16 bit double or quadruble the file size?
Title: Re: 8 or 16 bit
Post by: digitaldog on January 17, 2011, 01:09:17 pm
Quote from: Alan Klein on January 17, 2011, 01:01:16 pm
Would the 16 bit double or quadruble the file size?

Double.
Title: Re: 8 or 16 bit
Post by: ErikKaffehr on January 22, 2011, 02:11:41 am
Hi,

Yes, indeed. On the other hand 16 bits means in practice TIFF and that means fat files. When I did scanning I usually kept 16 bit TIFFs until final editing was done and saved as JPEG

Quote from: alain on January 16, 2011, 03:52:00 pm
...
The cost of 16-bit is so low it's a non issue. It would be foolish not to use.
...
Title: Re: 8 or 16 bit
Post by: alain on January 22, 2011, 06:40:46 am
Quote from: ErikKaffehr on January 22, 2011, 02:11:41 am
Hi,

Yes, indeed. On the other hand 16 bits means in practice TIFF and that means fat files. When I did scanning I usually kept 16 bit TIFFs until final editing was done and saved as JPEG

Eric

I just looked and a 2TB HDD was about 80 euro, so including 3 backupdrives where talking something in the range 240-300 euro for 1800000 megabyte usefull storage  or 9000 200 Megabyte 16-bit scans that are kept.  Scanning 9000 keepers is a large job.

When shooting RAW I just convert the keepers and those nr's aren't that high.  I only use jpeg for snapshots and some registrations where I won't do any editing after the raw-conversion.
Title: Re: 8 or 16 bit
Post by: DeeJay on January 22, 2011, 08:15:41 am
After posting this I worked on two of the same image. One 8bit and one 16bit.

The difference is substantial and apparent. When you play with contrast/colour, crunch the curves, the 8bit files breaks up, get choppy and bands long before the 16bit.

Having experimenting with it now I'll never retouch an 8bit file again.

Title: Re: 8 or 16 bit
Post by: DeeJay on January 22, 2011, 09:27:25 am
Quote from: digitaldog on January 17, 2011, 01:09:17 pm
Double.

It doubles the size, but am I right in saying it more than doubles save time? Something I'm convinced that I've observed.
Title: Re: 8 or 16 bit
Post by: alain on January 23, 2011, 08:52:55 am
Quote from: DeeJay on January 22, 2011, 09:27:25 am
It doubles the size, but am I right in saying it more than doubles save time? Something I'm convinced that I've observed.

That depends on you're hardware and image size.  Most HDD's will be configured to buffer the writes also and those buffers are something like 16,32,64 MB on modern drives.  You're computers says that it's ready before the image is really written to disk.
Title: Re: 8 or 16 bit
Post by: Philip Weber on February 14, 2011, 07:42:13 pm
This may be a dumb question but...

I'm working with 16-bit files out of Camera Raw, in Pro photo RGB. However, some Photoshop and Third Party Plug-ins require 8-bit files to do their work.

After I'm done, should I revert it back to a 16-bit file before flattening it as a TIFF or just leave it alone as an 8-bit file?

Thanks,
Phil
Title: Re: 8 or 16 bit
Post by: Schewe on February 15, 2011, 12:13:53 am
Quote from: Philip Weber on February 14, 2011, 07:42:13 pm
After I'm done, should I revert it back to a 16-bit file before flattening it as a TIFF or just leave it alone as an 8-bit file?

There is very, very little to be gained by re-converting to 16 bit. Going 16 to 8 bit is a one way operation– pretty much like a sex change operation...once you cut it off, you can never reattach what you cut off and expect it to work correctly.
Title: Re: 8 or 16 bit
Post by: stewarthemley on February 16, 2011, 05:15:42 am
Quote from: Schewe on February 15, 2011, 12:13:53 am
There is very, very little to be gained by re-converting to 16 bit. Going 16 to 8 bit is a one way operation– pretty much like a sex change operation...once you cut it off, you can never reattach what you cut off and expect it to work correctly.

Brilliant analogy. Love it.
Title: Re: 8 or 16 bit
Post by: stamper on February 16, 2011, 07:31:46 am
What happens if you are a female converter? ;D
Title: Re: 8 or 16 bit
Post by: bjanes on February 16, 2011, 10:25:20 am
Quote from: Guillermo Luijk on January 08, 2011, 04:24:45 pm
The point of switching to 16 bit is not making it real 16 bits, but making it more robust against strong pp.

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)

Switching to more than the original 8 bits will not improve the effective bitdepth of any resulting image, but will prevent posterization as in the extreme example above.

Regards

Guillermo,

A very nice demonstration, but we must remember that images taken with a real camera have noise, and the dithering produced by the noise will reduce posterization. Here is your 8 bit image with 2% noise applied in Photoshop using the Gaussian noise function.

8 bits per channel is usually sufficient for rendered reflection prints with a contrast ratio of up to 288:1 (8.2 stops) as in the ICC PRM (http://www.color.org/v4_prmg.xalter). This assumes that you got everything right with the camera exposure.  Prints will be with us for a long time, but as higher dynamic range display methods become more common, we will need more bit depth.

Greg Ward (http://www.anyhere.com/gward/hdrenc/hdr_encodings.html) has calculated that 8 bit sRGB has a DR of 1.6 orders of magnitude (5.31 stops), allowing for a maximum of a 5% difference at the low (shadow) end, which is quite generous, since the average person can detect even smaller changes. Of course, this discussion applies only to luminance, and when color images are involved, one needs even more bit depth as when going from sRGB to ProPhotoRGB.

From a scientific and human perceptual viewpoint, the use 16 bpc makes sense, and one might as well use a wide gamut space such as ProPhoto to make use of the colors captured by the camera.

Regards,

Bill