1 Correct answer
15 bit +1 was not an oversight but a deliberate design in order to have an odd number of integers and therefore give a middle value. Chris Cox who was an engineer at Adobe explained, in the past, that this was to speed up many of the blending calculations. Whether that reason is still valid with modern PC hardware is a point for debate, although some still complain of lags. It is though still at the core of Photoshop.
For visual work it makes no noticeable difference as the difference between
...
Explore related tutorials & articles
41
Replies
Perhaps a small example is helpful in understanding how Photoshop potentially destroys data in a 16bit integer file.
Here is a sample 16bit integer png with four shapes in it. The background value is set at a neutral 32768 for each RGB channel. Each shape is drawn at an integer 32769:
A "grey" shape drawn with RGB 32769,32769,32769
A "green" shape drawn with RGB 32768,32769,32768
A "blue" shape drawn with RGB 32768,32768,32769
A "red" shape drawn with RGB 32769,32768,32768
Open the file in Photoshop. Can you figure out what four shapes are hidden? 🙂
Btw, I wouldn't mind Photoshop throwing away data if the user would be warned against it. In my opinion it is rather unfortunate that users might rely on precise values maintained in their 16 bit integer files - be those 3d rendered, produced by scientific sensors, GIS depth maps, or whichever source - and the software fails to inform the user about it.
@TheDigitalDogYou might not have issues with this limit since you are a photographer, but in 3D, games, and scientific work it *is* an issue. I ran into this problem years ago while working on 3d projects. I suddenly experienced visible banding in 3d height maps where there was none before.
By @rayek.elfin
What issue?
Visible banding due to data/processing or visible banding due to the display path? The later is absolutely possible.
"Extraordinary claims require extraordinary evidence."-Carl Sagan
I await the evidence.
The simple fact remains that Photoshop reduces a 16bit image's precision to half - without informing the user. And the only reason why it was originally introduced by Chris Cox was for performance reasons back in the nineties.
Again, zero evidence was provided! As written and never addressed: Do tell us what capture device you have that produces true (full) 16-bit data. Or when you've seen 15+1 bit use, over the last 30 years is an issue?
Provide it, I'll ensure Adobe engineers at the top of the food chain see it. Balls in your court.
By @TheDigitalDog
@TheDigitalDogI do not presume to second guess your colour management and photography knowledge and experience, yet when you (again) display your misunderstanding as it relates to 3d height maps and (again) confuse banding in 3d mesh displacement resolution with visible image banding, I most respectfully ask you to inform yourself first about how height maps are used in 3d modeling and rendering apps. I am not talking about visual banding. I am talking about actual loss of resolution in the 3d displacement maps when 16bit greyscale depth maps are opened and saved in Photoshop.
I have created a simple gradient greyscale file that blends from 00000 to 65536 in steps of 1. When opened in Photoshop half the values are ignored by Photoshop and replaced with the previous step value. Which means 4 pixels are converted like so:
32768 - 32769 - 32770 - 32771 becomes 16384-16384-16385-16385
Half the data is lost.
This also happens with a RAW photo developed in Lightroom (full 16bit range) or a RAW developed in another image editor or developer: once it is opened in Photoshop intermediate values are not maintained. It is easy to check: develop a RAW file, save as a tiff, open in Photoshop, and save a tiff copy.
Open the original tiff in an image editor that does support true 16 bit integer files. I use PhotoLine, for example.
When I compare the original developed version in PhotoLine with the version I opened and saved in Photoshop: low and behold, all values are capped to odd ones. In the original values can be xxxxY where Y is either an even or odd number. In the Photoshop saved version Y can only be odd. Because all even values in the original 16bit range had to be downsampled to 32679 values from a range of 65535. That's how Photoshop solves it.
In short: all values in the original developed tiff are now limited to the 32769 range and values that differed 1bit in a channel are turned into equal values by Photoshop. There *is* a reduced colour fidelity. Data was lost.
Obviously visually it doesn't matter. Print can't capture this, nor screens or human eyes. But losing that fidelity in science, research, astro processing, or in visual effects (3d), games, what not - depending on how the files are used this is a very bad thing or at a minimum super inconvenient to work with: I am of the opinion that it is rather preferable to be avoided?
Or not? Should Photoshop be allowed to blind-side the unsuspecting user who expects that their original image data is left intact?
Shouldn't Photoshop at the very least inform the user about the incurred loss of data when opening an image file? I mean, Photoshop does warn us when we are about to save a layered file using a file format that doesn't support layers. A kind warning up-front would help to prevent potential frustration and misunderstanding. (It would have when I first encountered this issue many years ago!)
I do realize my statements in the previous post here were rather blunt and I apologize for that. Personally I have dealt with this (and still do) in my work. Because other artists have no idea about Photoshop messing with 16bit files that are generated from other software with a full 16bit range, and they then proceed to make some adjustments in PS, and losing half the data, and send it to us. It may not matter in one situation, but there are enough cases for us where it does matter.
And even in 32bit mode Photoshop's 16bit eyedropper tool still insists on pretending to be a 15bit ("+1") one. (Why?!) Nor is working with 32bit files always an option, as I've explained before. So a Photoshop user expecting the 16bit mode to not mess with their image data is problematic. And indeed, as suggested here, the only practical recourse is to switch to an alternative image editor. Because Photoshop is not enabling a proper 16bit workflow. And it destroys data without warning!
In short it's a major workflow hassle for those of us who do need a "real" 16 bit mode. At least, it was for me until I switched to other software for 16bit file editing. I mean, at the very least allow for true 16bit values when working in 32bit mode. But even in that image mode the 16bit values are limited to 32769? That is just odd in my opinion.
Still like Photoshop, though, and I work and deal with Photoshop on a daily basis, of course. I just wish it would be brought into the 21st century already. I do understand what Chris is saying, btw: at the time when I first hit this snag, I had a good discussion with him about it. He kept reiterating that it was done for code and memory performance reasons, which I understand. What I never understood till this day is why the devs haven't updated the 16 bit mode to allow the user to opt for a full 16bit range. My own musings are that it would affect large swathes of Photoshop's core code.
PS Photoshop is a bit inconsistent with how it interprets 16bit files. The values in the hidden message do pop up in Photoshop when converted to a TIFF file and imported - so PS seems to transform the colours when a 16bit TIFF is loaded. The values are still wrong, of course.
AdChoices
