Question about 16bit images in Photoshop

[Benhut1]

I have heard that Photoshop only uses 15bits of image data for all values except the highest value, for which it uses 16 bits. This means, that instead of the range of pixel values going from 0 to 0xFFFF, they go from 0 to 0x8000. At least that was the way it was in very early versions of Photoshop, like Photoshop 7. Has that changed? Does Photoshop now allow true 16-bit image data to be retained, so that when I save a full 16-bit image, it doesn't lose half of its available values? If so, which version of Photoshop first introduced the handling of true 16-bit image data? If not, then that means that 16-bit image processing is lossy in Photoshop, and that is not a desirable trait.

[D Fosse]

Right. Dave sums it up very nicely.

From my perspective, as a high-end photographer, the whole discussion can be condensed to how many angels can dance on the head of a pin - 16 or 15+1?

😉

[davescm]

'........how many angels can dance on the head of a pin ........?

That sounds like an idea for a future Something for the Weekend challenge 🙂

Dave

[PECourtejoie]

Dave, if someone, like in the videogame/scientific worlds examples given earlier, works in 32bits, but saves the image as 16 bits, would they get the entire precision they need, or 15+1?

[davescm]

'Dave, if someone, like in the videogame/scientific worlds examples given earlier, works in 32bits, but saves the image as 16 bits, would they get the entire precision they need, or 15+1?'

It's a good question Pierre and I don't know the answer. I would have to test outside of Photoshop as PS would open such an image in 15+1 regardless.

Dave

[D Fosse]

...but if they work in 32 bits and feel they need that, why would they save as 16 bit? It makes no sense. It's an artificial argument. Of course they wouldn't get the precision they need -?

It's as if I stored my master files as 8 bit, and then complained I didn't get the precision I needed.

What am I missing?

[PECourtejoie]

---

@D Fosse wrote:

...but if they work in 32 bits and feel they need that, why would they save as 16 bit? It makes no sense. It's an artificial argument. Of course they wouldn't get the precision they need -?

It's as if I stored my master files as 8 bit, and then complained I didn't get the precision I needed.

What am I missing?

---

The fact that they need 16 full bits, and not 32, and another user explained that they need 16 for precision and speed.(compared to 32)

[Lumigraphics]

I don't find this choice to be unacceptable at all. Engineering (and marketing) is about managing trade-offs. I don't like quite a few things in Adobe software but I also recognize that they serve a LOT of customers with differing needs.

I do take issue with your spinning this as "should be fixed" when many of us don't think its broken.

If its that much of a problem, 32-bit is available, or switch to different software. Surely someone doing high-end 3D and CGI work is using hefty workstations, right?

[TheDigitalDog]

---

@Lumigraphics wrote:

I do take issue with your spinning this as "should be fixed" when many of us don't think its broken.

Including Adobe and more importantly, ZERO proof of concept that anything IS broken.

As someone who's worked directly with Adobe engineers since version 2.5 of Photoshop, I kind of have an idea of how to get things fixed: first is proof something is broken. 

We've seen no proof. 

We've heard a lot of assumptions. Neither gets the team's attention, ever.  

[SAMEOBAKE]

As others have answered, I think the truth is that it is not that much of a concern in visible photography and video. Also, the production process in recent years often involves the use of floating point numbers.

On the other hand, if the values that need to be accurate are saved and used as images, it is also true that simply passing the software can degrade the image. I have encountered this problem several times in my work.

As a person who writes code, I understand that we made this specification at the time of development amidst performance and various other issues, and I do not expect Adobe to fix these issues.

I have tested other software and were able to find one that can process 16-bit images correctly, so we plan to move to that.

[davescm]

'if the values that need to be accurate are saved and used as images'

Can you describe your use case where retaining full 16 bit accuracy in images is needed and where 15 bit+1 makes a real world difference? Certainly not in photography which tends to start in camera as 14 bit raw data, so 15bit+1 is more than enough for adjustments.  Display on screens is 10 bit at best, and 6 bit +dither at worst. Tests on printed output show no visible difference between 16 bit and 8 bit sent to printer.

I can think of one case in my workflow which is for 3D height maps. In that case I use 32 bit floating point.

Dave

[D Fosse]

This comes up with clockwork regularity, usually from people who make it a point to find "flaws" in Photoshop. Not one of them has ever demonstrated why this should have any practical significance whatsoever.

[rayek.elfin]

Not from a point of view of photography. And files are noticeably smaller when saved as Photoshop 16bit.

It is inconvenient for displacement maps in 3d work, however. Half the resolution is lost, and 32bit is not an option for various reasons. So only possible via other alternatives like Affinity, Gimp, or PhotoLine.

From a computer science perspective it is problematic, however. And Photoshop is used in those fields as well. The expectation is that a 16bit image saved from Photoshop is indeed a full range 16bit image. but that is of course not the case, which can have unexpected results when analysing that image.

Here is a quoted post made by a programmer who works on a medical application:

First [..] this is not a photography but a computer science question. Second, the basis for the photography and artistic things you are referring to is strict maths and computer science, and there one bit makes a world of a difference.

With 16 bits the maximum value will be twice as much as with 32 bits. Just imagine you apply a logarithmic scale after normalising to [0-1]. You might end up unexpectedly getting negative values.

I just came across this problem when trying to find out the maximum intensity in an x-ray projection for a CT reconstruction. And yes, the values Photoshop gives are not precise; OpenCV gives me app. the double. This is a big problem if you're seeking to precisely analyse an image in Photoshop.

https://www.dpreview.com/forums/thread/3847350#forum-post-55855598

The average user and most expert users are not affected by it, of course. That doesn't mean it wouldn't be good to have a full 16bit mode in Photoshop for outlier cases.

Btw, I am happy to see that the 8bit display pyramid problem when viewing 16bit files when zoomed out was finally fixed in the latest release.

[TheDigitalDog]

The average user and most expert users are not affected by it, of course. That doesn't mean it wouldn't be good to have a full 16bit mode in Photoshop for outlier cases.

By @rayek.elfin

Start here:

https://community.adobe.com/t5/photoshop-ecosystem-ideas/how-do-i-write-a-feature-request/idi-p/1238...

[AllenY.NK]

Here to support fellow 3D/VFX/game artists and computer science users. @rayek.elfin's concern and use cases are legit. Rayek's being professional and well-mannered (with some frustration), but many other so called profesional photographers/community experts were just failed to see the real points and keep bullying Rayek. If you don't have the usage and/or don't understand the use of full 16-bit range, doesn't mean it is not importand or not legit. Albeit you are a very experienced Photoshop user and long time member in the commnunity.

Photoshop started as a profesional photography software, but over the decades it is already grown into an image processor with much broader usage, well outside of field of photography. Many other users in the thread (like johnm42377455 & Alex25613730lde7) already pointed out the usage involved with the full 16-bit range. So, please look into those usage and understand them before you dismiss other people's point just because your own ignorance.

Height maps/normal Maps/computer science images all have legit and important uses of the full range of 16-bit. There are performance differences between 16-bit and 32-bit float and many industries have legit uses of 16-bit image files. Tell people to use other software if they need true 16-bit is just to overlook the importance of it. Plus, Adobe already acquired the Substance suite and integrated into their product line, also tries to make them all work together with other non-Substance products. It is more important than ever to understand and address the lack of true 16-bit mode.

[davescm]

I see no bullying here, nor would I call people ignorant.

What I do see is advice on how to make a feature request which will mean raising it under 'ideas' and outlining in detail use cases that would benefit from 16 bit integer but not be served by 32 bit float. Then getting sufficient votes to put it on the radar of Adobe's development team. This discussion thread is not such a request.

Dave

[TheDigitalDog]

I see no bullying here, nor would I call people ignorant.

What I do see is advice on how to make a feature request which will mean raising it under 'ideas' and outlining in detail use cases that would benefit from 16 bit integer but not be served by 32 bit float. Then getting sufficient votes to put it on the radar of Adobe's development team. This discussion thread is not such a request.

Dave

By @davescm

Exactly! Since no feature request exists here what we have is four years of a discussion going nowhere. Adobe engineering isn't paying attention and some wonder why. So again.

Start here:

https://community.adobe.com/t5/photoshop-ecosystem-ideas/how-do-i-write-a-feature-request/idi-p/1238...

[Lumigraphics]

My wild guess is that this would be a huge architectural change that is not on Adobe's development roadmap because of the work involved and limited benefit to most customers.