fredmiranda.com

Moderated by: Fred Miranda

New fredmiranda.com Mobile Site

New Feature: SMS Notification alert

New Feature: Buy & Sell Watchlist

FM Forums | Post-processing & Printing | Join Upload & Sell

1 2

4 end

Archive 2025 · 16 bit vs. 14 bit ...

Jack Flesher
Offline
• • • •
Upload & Sell: On

p.3 #1 · 16 bit vs. 14 bit ...

RustyBug wrote:
+1 that color is where I find 14 bit to tap out and the colors begin to breakdown / artifact when pushing them hard ... at the edges, as you suggest.

While others have been ardent at suggesting that there is no difference, it's good to hear that someone recognizes that there is a realm where the difference is helpful.

I'll probably follow up with some direct contacts to Hasselblad and Fuji to get their take on this, too. Armchair forum dissension, notwithstanding ... I do want to further understand the implications / effects. I sorta know where my 14 bit files...Show more →

Yes 😎. Where I see it mostly in my personal imaging now, is in fluorescent colors on signs while doing basic travel photography, and occasionally in super intense sunrises or sunsets. I personally am not in the habit of heavily pushing images in post, or going as heavy-handed on saturation as some do, so for me these occasional edge cases are virtually irrelevant for *my* work and I live happily with my 14-bit camera.

Nevertheless I respect that others notice it frequently enough it bothers them � and I used to be one

� and at the same time respect others either never have seen it or at least never noticed it and also are fine living without it. But where I draw the line is accepting and defending that it exists� IOW, it�s real, but it�s relevance is up to the individual. 👍

Apr 25, 2025 at 10:50 AM

JimKasson
Offline
• • •
Upload & Sell: On

p.3 #2 · 16 bit vs. 14 bit ...

gdanmitchell wrote:
... when the destination format only supports 8 bits, there are still good reasons to have more than 8 bits of data available at prior stages.

You are talking about apples and oranges. None of the 8-bit interchange formats that are popular use gamma equals unity, and with good reason.

Apr 25, 2025 at 12:00 PM

Jack Flesher
Offline
• • • •
Upload & Sell: On

p.3 #3 · 16 bit vs. 14 bit ...

JimKasson wrote:
You are talking about apples and oranges. None of the 8-bit interchange formats that are popular use gamma equals unity, and with good reason.

Maybe I'm missing something. I believe in this discussion we're talking about digital images specifically, and why extra bit-depth *can* help while processing out an image intended for some specific 8-bit display medium. IMHO this is the apple. And yes, it is OUR job to make sure we render the 16, 14 or any 8+ bit image down to the 8-bit image using the proper protocols for the target medium. What it seems like you are addressing are the standardized ANSI hand-shakes between/among 8-bit devices, and if so, that is relatively an orange for this discussion.

Any gamma correction or profile trimming is handled during conversion based on how we instructed it via rendering intent and black-point handling at that time -- after that we ship it downstream and it's up to the devices in that chain to each do their job according to a standard...

Apr 25, 2025 at 01:35 PM

gdanmitchell
Offline
• • • • • • •
Upload & Sell: Off

p.3 #4 · 16 bit vs. 14 bit ...

JimKasson wrote:
You are talking about apples and oranges. None of the 8-bit interchange formats that are popular use gamma equals unity, and with good reason.

II think you missed my point.

Apr 25, 2025 at 04:00 PM

RustyBug
Offline
• • • • • • •
Upload & Sell: On

p.3 #5 · 16 bit vs. 14 bit ...

Jack Flesher wrote:
IOW, it�s real, but it�s relevance is up to the individual. 👍

Apr 25, 2025 at 06:06 PM

JimKasson
Offline
• • •
Upload & Sell: On

p.3 #6 · 16 bit vs. 14 bit ...

gdanmitchell wrote:
II think you missed my point.

How's that?

Apr 26, 2025 at 12:13 PM

JimKasson
Offline
• • •
Upload & Sell: On

p.3 #7 · 16 bit vs. 14 bit ...

Jack Flesher wrote:
Maybe I'm missing something. I believe in this discussion we're talking about digital images specifically, and why extra bit-depth *can* help while processing out an image intended for some specific 8-bit display medium. IMHO this is the apple. And yes, it is OUR job to make sure we render the 16, 14 or any 8+ bit image down to the 8-bit image using the proper protocols for the target medium. What it seems like you are addressing are the standardized ANSI hand-shakes between/among 8-bit devices, and if so, that is relatively an orange for this discussion.

Any gamma correction...Show more →

The tone curve used affects the quantizing noise, the banding, the color accuracy, and many other things. It makes no sense to conflate linear representations with nonlinear ones based on the precision of the quantization. You need to look deeper than that. For example, the quantization noise floor of an 8-bit file with the sRGB tone curve is less than a linear file quantized at 11 bits.

Apr 26, 2025 at 12:17 PM

dclark
Offline
• • • • •
Upload & Sell: On

p.3 #8 · 16 bit vs. 14 bit ...

It's great to see Jim Kasson, a person with deep understanding of image science, participating in the FM forums. I hope he continues to contribute.

Apr 29, 2025 at 10:32 PM

ruthenium
Offline
• • • •
Upload & Sell: On

p.3 #9 · 16 bit vs. 14 bit ...

I have been looking again at the engineering dynamic range data for GFX100 II and GFX100S II from Jim (https://blog.kasson.com/gfx-100s-ii/fujifilm-gfx-100s-ii-edr/) and also at the photographic dynamic range data for GFX100 II from Photons to Photos.
Jim noted "16 bit precision doesn�t appreciably improve the EDR." Indeed, there is very little difference, and the numbers at key ISO 80 and 500 look practically the same.
On Photons to Photos, there is a single graph for PDR. I don't see any comments on whether the data were obtained at 16 or 14 bit precision. I guess there wasn't any difference as well?
These observations leave me wondering, how the 16 bit precision makes some difference to photos? If it does indeed (that I am not questioning here), but the effect is not captured in the EDR/PDR numbers, then what is it that makes the difference and why this effect is not seen from the EDR/PDR numbers?
I am merely interested in developing a proper understanding of the numbers and what these numbers do or don't tell in practice. I know that what I am asking might be trivial; nevertheless, I would appreciate it if someone could answer the question for me - thanks!

Apr 29, 2025 at 11:39 PM

RustyBug
Offline
• • • • • • •
Upload & Sell: On

p.3 #10 · 16 bit vs. 14 bit ...

ruthenium wrote:
I have been looking again at the engineering dynamic range data for GFX100 II and GFX100S II from Jim (https://blog.kasson.com/gfx-100s-ii/fujifilm-gfx-100s-ii-edr/) and also at the photographic dynamic range data for GFX100 II from Photons to Photos.
Jim noted "16 bit precision doesn�t appreciably improve the EDR." Indeed, there is very little difference, and the numbers at key ISO 80 and 500 look practically the same.
On Photons to Photos, there is a single graph for PDR. I don't see any comments on whether the data were obtained at 16 or 14 bit precision. I guess there wasn't any difference as well?
These observations...Show more →

I can't directly answer the reason why the test data doesn't show the difference, since I'm not technically savvy about the underlying aspects of the testing.

I will simply propose that the testing doesn't apply the same amounts of "stacked math" that comes from applying a multitude of mathematical operations, particularly where aggressive application of exponential operators (i.e. curves, etc.) are compounded with other aggressive operations (dehaze / clarity, etc.) and sharpened, et al. Because (good) testing protocols are standardized, the extremes of (torture) testing aren't likely to be part of the standardized testing approach.

Which, of course takes us back to the point that only in extreme (vs. routine) use, does the difference reveal itself in an 8 or 10 bit presentation.

Apr 30, 2025 at 06:37 AM

Jack Flesher
Offline
• • • •
Upload & Sell: On

p.3 #11 · 16 bit vs. 14 bit ...

JimKasson wrote:
The tone curve used affects the quantizing noise, the banding, the color accuracy, and many other things. It makes no sense to conflate linear representations with nonlinear ones based on the precision of the quantization. You need to look deeper than that. For example, the quantization noise floor of an 8-bit file with the sRGB tone curve is less than a linear file quantized at 11 bits.

And why is that? Because the 11 bit has more data available. Are we falsely pushing non-visible or less distinct hues into more visible/distinct hues? Yes, or at least probably. But that�s our choice as artists. We�re close to this discussion devolving into how the human eye-brain renders transmissive vs reflective color; and that magenta isn�t a real color, but any way you slice it, more information is better when squishing or stretching colors so they appropriately fit into a chosen output medium to look the way we want them to.

However as I said earlier, I accept the difference between 14-bit and 16-bit is virtually irrelevant in most cases except where extreme hues and saturation exist, and for the most part 14-bit is more than adequate for most imaging applications. In fact, I regularly throw away the end bits to force harder black and white into my images�

Apr 30, 2025 at 08:50 AM

ruthenium
Offline
• • • •
Upload & Sell: On

p.3 #12 · 16 bit vs. 14 bit ...

RustyBug wrote:
I can't directly answer the reason why the test data doesn't show the difference, since I'm not technically savvy about the underlying aspects of the testing.

I will simply propose that the testing doesn't apply the same amounts of "stacked math" that comes from applying a multitude of mathematical operations, particularly where aggressive application of exponential operators (i.e. curves, etc.) are compounded with other aggressive operations (dehaze / clarity, etc.) and sharpened, et al. Because (good) testing protocols are standardized, the extremes of (torture) testing aren't likely to be part of the standardized testing approach.

Which, of course takes us back to...Show more →

After more thinking, I believe that my problem/confusion could be about misinterpreting 14 bit or 16 bit as dynamic ranges of a sort whereas these mean precisions.
A dynamic range such as EDR or PDR is a ratio. The numerator is the same - the largest value in a range but the denominator is defined differently. In the simplest definition of EDR, the denominator is the number corresponding to the point where S/N = 1. Obviously, the ratio that gives the dynamic range is influenced by the numerical precision of the point where S/N = 1 but only until the calculated log base 2 of the ratio stops changing when rounded to two decimal places. The observation of practically the same EDR at 14 and 16 bit precision indicates that the added numerical accuracy of the later no longer perceptibly changes the calculation of the ratio of the largest value in the range to the numerical value of the point when S/N =1.
The denominator of the photographic dynamic range (PDR) is defined in a clever way to allow a direct comparison of cameras with sensors of different sizes or images obtained from the same camera while cropping, all assuming that these are scaled(printed) to be of the same size and are viewed from the same distance. I might be wrong but it seems that PDR can be connected with the photographic equivalence. That is two images obtained with camera systems utilizing equivalent sensor technology should have the same PDR when the images are equivalent. In real life, the PDR values can be expected to be slightly different, given that cameras from different brands do have some differences in the underlying technology.
Here is an example: I am using FF and MFT cameras (A1, OM-1 II, G9 II) and this thread made me interested in possibly adding a GFX100S II with the "standard" zoom GF32-64mm F4. One question that can be asked is whether this camera system should give me a better low-light performance than an MFT camera system that I own? The highest native ISO on the GFX100S II is 12,800. The crop factor to MFT is 2.53 or when squared it is 6.41 (the ratio of the diagonals is 54.78/21.64mm). To obtain an equivalent ISO on my G9 II, I have to set it to 12800/6.41 = ISO 2000 while using a lens that has an f/ratio of 4/2.53 = f/1.6. Most of my use of the Panasonic Lumix G9II camera is with the Leica 10-25 f/1.7 zoom (FF equivalent: 20 - 50 mm), and f/1.7 is reasonably close to the calculated f/1.6 to ignore the difference. Thus, when used side-by-side the G9II with the 10-25mm zoom at f/1.7 set to ISO 2000 and the GFX100S with the 32-64 zoom at f/4 set to ISO12,800 are expected to produce nearly equivalent images at an equivalent FL. The PDR is 6.58 for the former and 6.37 for the latter - very close, as I expected.
My present understanding of the subject of this thread (that can change if I should be corrected by the experts): 16 vs 14 bit makes little difference for the dynamic range but it does increase the precision of the R, G, and B values that collectively give colors of different tonality. More precision gives more subtle tonal variations. These may or may not be needed and may or may not be visible depending on how the product image is presented and viewed.

Apr 30, 2025 at 09:23 PM

JimKasson
Offline
• • •
Upload & Sell: On

p.3 #13 · 16 bit vs. 14 bit ...

Jack Flesher wrote:
And why is that? Because the 11 bit has more data available. Are we falsely pushing non-visible or less distinct hues into more visible/distinct hues? Yes, or at least probably. But that�s our choice as artists. We�re close to this discussion devolving into how the human eye-brain renders transmissive vs reflective color; and that magenta isn�t a real color, but any way you slice it, more information is better when squishing or stretching colors so they appropriately fit into a chosen output medium to look the way we want them to.

However as I said earlier, I accept the...Show more →

Do you really not understand how the tone curve affects the quantizing limited dynamic range? Or is that a rhetorical question?

Once the signal is swamped by the noise, there is no point in greater quantizing precision.

May 01, 2025 at 11:54 AM

Jack Flesher
Offline
• • • •
Upload & Sell: On

p.3 #14 · 16 bit vs. 14 bit ...

JimKasson wrote:
Do you really not understand how the tone curve affects the quantizing limited dynamic range? Or is that a rhetorical question?

Once the signal is swamped by the noise, there is no point in greater quantizing precision.

Yes, of course I do. By definition, below a certain noise floor, assuming that's what you mean by "swamped," is what defines the lower limit of DR to begin with, and obviously any curve we apply afterward shrinks DR. What we are discussing here are colors that are within usable DR AND translatable within colorspace chosen to display them. Hence, if I destroy DR with an aggressive tone curve in a 14-bit image, then yes, 16-bit data may be irrelevant. But it isn't only pure highlights and pure shadows that we're discussing, there are other high and low edge colors that can be saved even with some curve applied, especially if it's carefully and properly applied. And as such, if I destroy some of the 16-bit data say down to a real value of anything over 14-bit, then 16-bit origin files become a useful option... And in *MY* experience, there have been situations where *I* saw a real benefit of starting with said 16-bit file.

So my question to you is, have you ever done critical color product photography or fine art reproduction work under controlled studio lighting, especially subjects that include super-saturated and/or fluorescing colors? If so, then for sure you would know what I say above is real. By your continued adherence to stressing quantizing limits, I have to assume you've never actually done any of that type of imaging, or simply have never worked a file with a specific goal to preserve a critical color, and as such may be the reason you can't comprehend the usefulness of 16-bit capture. Joe Holmes and I actually discussed this years ago when the early Phase and Hassy backs achieved over 14-bit linear capability.

Once again, what I am talking about are indeed edge use cases, and I would agree that 16-bit is unnecessary for most "normal" photography. But outside that norm there are situations where 16-bit --or more accurately, anything more than 14-bit-- can and does help.

May 01, 2025 at 12:26 PM

JimKasson
Offline
• • •
Upload & Sell: On

p.3 #15 · 16 bit vs. 14 bit ...

Jack Flesher wrote:
Yes, of course I do. By definition, below a certain noise floor, assuming that's what you mean by "swamped," is what defines the lower limit of DR to begin with, and obviously any curve we apply afterward shrinks DR. What we are discussing here are colors that are within usable DR AND translatable within colorspace chosen to display them. Hence, if I destroy DR with an aggressive tone curve in a 14-bit image, then yes, 16-bit data may be irrelevant. But it isn't only pure highlights and pure shadows that we're discussing, there are other high and low edge...Show more →

I have worked on cultural preservation. I have seen no advantage of 16 bit precision in those circumstances. WRT to the rest of the tone curve, you are concentrating on read noise and ignoring shot noise.

May 01, 2025 at 01:15 PM

Jack Flesher
Offline
• • • •
Upload & Sell: On

p.3 #16 · 16 bit vs. 14 bit ...

JimKasson wrote:
I have worked on cultural preservation. I have seen no advantage of 16 bit precision in those circumstances. WRT to the rest of the tone curve, you are concentrating on read noise and ignoring shot noise.

I accept you haven�t seen it. I have seen it, Joe Holmes has seen it, and he�s probably forgotten more about digital color science than you and I combined know about it.

May 01, 2025 at 02:16 PM

RoamingScott
Offline
• • • • • •
Upload & Sell: Off

p.3 #17 · 16 bit vs. 14 bit ...

This thread is everything I knew it could be.

May 01, 2025 at 02:17 PM

JimKasson
Offline
• • •
Upload & Sell: On

p.3 #18 · 16 bit vs. 14 bit ...

Jack Flesher wrote:
I accept you haven�t seen it. I have seen it, Joe Holmes has seen it, and he�s probably forgotten more about digital color science than you and I combined know about it.

Papers:

Kasson, J.M., �Efficient, Chromaticity-Preserving Sharpening for RGB Images,� Device-Independent Color Imaging, Walowit, E., Editor, SPIE vol. 2414, pp. 134-145 (1995).

Kasson, J.M., �Efficient, Chromaticity-Preserving Midtone Correction for RGB Images,� Second Color Imaging Conference, Scottsdale, AZ, November 15-18, 1994.

Kasson, J. M., �Tetrahedral Interpolation Algorithm Accuracy,� Device-Independent Color Imaging, Walowit, E., Editor, SPIE vol. 2170 (1994).

Kasson, J. M., Plouffe, W.E., and Nin, S. I., �A tetrahedral interpolation technique for color space conversion,� Device-Independent Color Imaging and Imaging Systems Integration, Motta, R. J., and Berberian, H. A., Editors, SPIE vol. 1909, pp 127-138 (1993).

Nin, S. I., Kasson, J. M., and Plouffe, W., �Printing CIELAB images on a CMYK printer using trilinear interpolation,� Color Hard Copy and Graphic Arts, Bares, J., Editor, SPIE vol. 1670 (1992).

Plouffe, W. Kasson, J.M., Easy-to-Compute Non-Linearities for Efficient Encoding of Color, Society for Information Display International Symposium, Digest of Technical Papers, Volume XXII, May 1991, pp 814-816.

Kasson, J.M. and Plouffe, W., �Subsampled Device-Independent Interchange Color Spaces,� Image Handling and Reproduction Systems Integration, SPIE vol 1460, pp 11-19, 1991.

Kasson, J.M., Color Science for Device-Independent Color Reproduction, Society for Information Display Conference, Las Vegas, NV, May 1990.

Kasson, J.M. and Plouffe, W., Requirements for Computer Interchange Color Spaces, SPSE/SPIE Electronic Imaging Conference, Santa Clara, CA, February 1990.

Publications:

Kasson, J.M., Nin, S.I., Plouffe, W.E., and Hafner, J.L., �Performing Color Space Conversions with Three-Dimensional Linear Interpolation, Journal of Electronic Imaging, vol. 4, July 1995, pp. 226-249.

Kasson, J.M., and Plouffe, W.E., �An Analysis of Selected Computer Interchange Color Spaces�, ACM Transactions on Graphics, vol. 11, no. 4, October, 1992, pp. 373-405.

Patents:

Kasson, J.M., �Computationally efficient low-artifact system for spatially filtering digital color images,� USA 5,793,885, issued August 11, 1998.

Kasson, J.M., �Method and Apparatus for Tone Correction of a Digital Color Image with Preservation of the Chromaticity of the Image,� USA. 5,774,112, issued June 30, 1998.

Kasson, J.M., and Plouffe, W.E., Pryor, D., Nin, S.I., �Function Approximation Using a Centered Cubic Packing with Tetragonal Disphenoid Extraction,� USA 5,751,926, issued May 12, 1998.

Edgar, A., and Kasson, J.M., �Automatic Cross Color Elimination,� USA. No. 5,509,086, issued Apr 16, 1996.

Kasson, J.M., �Method and Apparatus for Interactively Indicating Image Boundaries in Digital Image Cropping,� USA. 5,473,740, issued December 5, 1995.

Kasson, J.M., �Color Image Gamut-Mapping System with Chroma Enhancement at Human-Insensitive Spatial Frequencies,� USA. 5,450,216, issued September 12, 1995.

Kasson, J.M., and Plouffe, W.E., �Tetrahedron/Octahedron Packing and Tetrahedron Extraction for Function Approximation,� USA. 5,390,035, issued February 14, 1995.

May 01, 2025 at 05:08 PM

JimKasson
Offline
• • •
Upload & Sell: On

p.3 #19 · 16 bit vs. 14 bit ...

Jack Flesher wrote:
I have seen it, Joe Holmes has seen it, and he�s probably forgotten more about digital color science than you and I combined know about it.

I asked Joe about this. Here's what he said: "I record only 14-bit files with my GFX 100S."

Jim

May 07, 2025 at 11:13 AM

Jack Flesher
Offline
• • • •
Upload & Sell: On

p.3 #20 · 16 bit vs. 14 bit ...

JimKasson wrote:
I asked Joe about this. Here's what he said: "I record only 14-bit files with my GFX 100S."

Jim

Okay, fair enough.

May 07, 2025 at 05:13 PM

1 2 3 4 end
FM Forums \| Post-processing & Printing \| Join Upload & Sell 1 2 3 4 end