32bit image in Ps CC

[Tonyhaz]

After running DSS, and I get the auto save image ( newbie ) I open it in photoshop CC, but I noticed I was unable to use the curves tool along with many others, the file is 32bit, I changed it to 16bit and worked ok, why won’t it work in 32bit ? Surly I’m loosing image quality if I reduce it. I’ve added a pic of what is available in 32bit, even after I duplicated it don’t let me. 

[sharkmelley]

Photoshop 32-bit mode is a very specialised mode and is quite restricted in terms of supported functionality.  As you have discovered there are limited operations that can be performed on the data.

However, you need to be aware that because the Photoshop 32-bit mode is specialised, it is probably not doing what you expect.  Photoshop assumes a 32-bit file is linear data.  So as soon as you open a 32-bit file in Photoshop it decides to apply the relevant colour space gamma to the displayed image - this will be a gamma of around 2.2 depending on which colour space you are working in i.e. sRGB, AdobeRGB or whatever.  However, the RGB values it reports at the cursor position are the original linear values it reads from the file and all the operations you perform on the image (e.g. background subtraction) are performed using the linear values.  If you then change mode to 16-bits using the "Exposure and Gamma" method with default values in the HDR Toning dialog you will notice that the displayed image does not change but the RGB values it reports are now the completely different gamma adjusted values.  You don't lose any image quality converting from 32-bits to 16-bits because the 16-bit gamma adjusted values have a greater dynamic range than the original 32-bit linear values.

For completeness I should clarify that when you open a 16-bit or 8-bit image in Photoshop then it assumes that the values it reads from the file are already in the relevant colour space with the gamma already applied so it does not apply the colour space gamma to the displayed image.  This is the case unless you have deliberately specified an alternative specific ICC colour profile in the file.

A one line summary is that when you open a 32-bit linear file in Photoshop, it automatically displays a gamma transformed version of your data.  You need to decide whether or not this is what you want.

You can easily prove this to yourself by saving the same stacked data from DSS as a 16-bit file or as a 32-bit file.  When you open these two files in Photoshop they will look completely different because Photoshop applies gamma to one but not the other.  Unless you are well acquainted with colour management, colour spaces, gamma transfer curves etc then the pragmatic solution is simply to go with what you think looks best ?

Mark

[tooth_dr]

13 minutes ago, sharkmelley said:

Photoshop 32-bit mode is a very specialised mode and is quite restricted in terms of supported functionality.  As you have discovered there are limited operations that can be performed on the data.

However, you need to be aware that because the Photoshop 32-bit mode is specialised, it is probably not doing what you expect.  Photoshop assumes a 32-bit file is linear data.  So as soon as you open a 32-bit file in Photoshop it decides to apply the relevant colour space gamma to the displayed image - this will be a gamma of around 2.2 depending on which colour space you are working in i.e. sRGB, AdobeRGB or whatever.  However, the RGB values it reports at the cursor position are the original linear values it reads from the file and all the operations you perform on the image (e.g. background subtraction) are performed using the linear values.  If you then change mode to 16-bits using the "Exposure and Gamma" method with default values in the HDR Toning dialog you will notice that the displayed image does not change but the RGB values it reports are now the completely different gamma adjusted values.  You don't lose any image quality converting from 32-bits to 16-bits because the 16-bit gamma adjusted values have a greater dynamic range than the original 32-bit linear values.

For completeness I should clarify that when you open a 16-bit or 8-bit image in Photoshop then it assumes that the values it reads from the file are already in the relevant colour space with the gamma already applied so it does not apply the colour space gamma to the displayed image.  This is the case unless you have deliberately specified an alternative specific ICC colour profile in the file.

A one line summary is that when you open a 32-bit linear file in Photoshop, it automatically displays a gamma transformed version of your data.  You need to decide whether or not this is what you want.

You can easily prove this to yourself by saving the same stacked data from DSS as a 16-bit file or as a 32-bit file.  When you open these two files in Photoshop they will look completely different because Photoshop applies gamma to one but not the other.

Mark

 

Recently I opened a 32 bit tiff saved in APP and it looked completely different and wrong compared to the 16 bit version I opened just prior to it.  Thanks for this explanation ??

[Davey-T]

And some things PS only does in 8 bit.

If it's only going to be viewed on a PC screen it doesn't matter much.

Dave

[Merlin66]

As an alternative the highly capable PaintShopPro comes in 32 and 64 bit.......certainly come a long way from its sketching beginnings.

https://www.paintshoppro.com/en/

Does 99% of all the stuff PS does for a fraction of the cost!

 

 

[sharkmelley]

4 hours ago, Merlin66 said:

As an alternative the highly capable PaintShopPro comes in 32 and 64 bit.......certainly come a long way from its sketching beginnings.

https://www.paintshoppro.com/en/

Does 99% of all the stuff PS does for a fraction of the cost!

 

But what, specifically, does PaintShopPro do when a 32-bit image is opened?  Does it behave in the same manner as Photoshop?

In other words, if you save the same stacked data from DSS as a 16-bit file or as a 32-bit file and open these two files in PaintShopPro  will they look completely different because PaintShopPro applies the colour space gamma transfer curve to one but not the other?

Mark

[Merlin66]

Mark,

I believe it would meet your needs....

Download the demo version and try it out.

 

[dannybgoode]

7 hours ago, Merlin66 said:

As an alternative the highly capable PaintShopPro comes in 32 and 64 bit.......certainly come a long way from its sketching beginnings.

https://www.paintshoppro.com/en/

Does 99% of all the stuff PS does for a fraction of the cost!

 

 

The fact that Paint Shop Pro comes in a version for 32 bit and 64 bit operating systems is completely unrelated to how it does or whether it even can handle 32 bit images. As it happens (and I’ll need to double check) I don’t think it can handle 32 bit images at all. 

@sharkmelley gives the best explanation of what the implications are for 16 or 32 bit images. 

As an aside Affinity Photo is only £38 at the moment and beats PSP in every area. 

[Merlin66]

I think you and others who are more active in AP are in a better position to verify the capabilities.

I’m of the opinion that if your needs can be met in PS then PSP will also do the job.

A quick run through the demo version will confirm or otherwise........

[happy-kat]

In general plugins can be run in both but PSP can't make use of PS action scripts.

[billyharris72]

9 hours ago, Merlin66 said:

Does 99% of all the stuff PS does for a fraction of the cost!

Sounds good. Other option worth exploring is the new version of Gimp, which has 32 bit support (after years of being 8 bit only) and a great array of tools (comparable to though not as user friendly as PS, and without the PS astro plugins).

You could try using Gimp for the initial stretch and then going back to PS for the rest of the workflow if you don't get on with it. Nothing to lose given that it's free.

Billy.

[Thalestris24]

I think the dss autosave tiff is a 32-bit floating point file. You can open it in the latest version of Gimp and/or you can optionally save the dss stacked output in a different format if you wish. Gimp also opens fits files which is quite useful. Maybe PS does also but I don't use it so don't know

Louise

[sharkmelley]

For what it's worth I downloaded GIMP and opened 16-bit and 32-bit linear TIFFs (with no embedded ICC colour profile) saved from the same stacked data.  The two files appeared identical on the screen.  In other words, GIMP does not perform the same weird 32-bit trick that Photoshop does.

Mark

[Merlin66]

Can you try the same exercise with PSP??

 

[sharkmelley]

PSP 2019 makes no mention of 32-bit files.  In any case I installed the PSP trial but it was unable to open the 32-bit TIFF:  "Unable top open the file.  Please verify that the file is valid".

Mark

 

[Marci]

PSP does NOT support 32bit-depth image formats - only up to 16bit-depth. See http://help.corel.com/paintshop-pro/v20/main/en/documentation/index.html#page/Corel_PaintShop_Pro/Understanding_color_depth.html

 

Colour bit-depth of an image is _completely_ unrelated to memory addressing bitsize of the targeted software architecture (i.e.: 32bit software/OS/hardware v 64bit software/OS/hardware).

 

That said, your camera won't be producing 32bit subs - it'll be producing 8, 12 or 14bit (iirc for most common sensors, 16bit if it's an expensive astro-specific scientific camera), so it's pointless sticking with DSS's default behaviour of saving 32bit TIFs. You can't invent data (the additional 16bits of color depth) that isn't there & recorded by the sensor to begin with. Set DSS to save as 16bit, then use any software supporting 16bit color depth images (PS / PSP / Gimp / Affinity etc, regardless of 32bit / 64bit architecture).

pixInsight & Maxim (& probably other scientific image processors, as opposed to photographic visual image editors) will handle 32bit-depth image formats, but bear in mind any data beyond the bit-depth of the camera sensor is invented/backfilled by mathematical algorithm.

 

[sharkmelley]

56 minutes ago, Marci said:

That said, your camera won't be producing 32bit subs - it'll be producing 8, 12 or 14bit (iirc for most common sensors, 16bit if it's an expensive astro-specific scientific camera), so it's pointless sticking with DSS's default behaviour of saving 32bit TIFs. You can't invent data (the additional 16bits of color depth) that isn't there & recorded by the sensor to begin with.

 

I disagree. 

If my camera has a 14-bit dynamic range and I stack 256 (linear data) subs in DSS then that potentially gives me a 22-bit dynamic range in the stack.  If I save that as a 16-bit (linear data) file then it potentially truncates the dynamic range of my stack.  The authors of programs like DSS and PixInsight allow 32-bit files for a good reason.

Mark

[Tonyhaz]

Thanks for all reply’s, 16 bit it is from now on. 

[Merlin66]

Hmmmm

I. Must be missing something here...

I use an ASI 1600 camera for solar....regularly take 1000 frames (up to a 3Gb SER file), the frames are said to be 12 bit files....when they are summed the final result is still a 12 bit file - not a 48 bit file??????

[sharkmelley]

1 hour ago, Merlin66 said:

Hmmmm

I. Must be missing something here...

I use an ASI 1600 camera for solar....regularly take 1000 frames (up to a 3Gb SER file), the frames are said to be 12 bit files....when they are summed the final result is still a 12 bit file - not a 48 bit file??????

Let's take your example and round it up to 1024 frames to make the maths easier.

A single frame is 12-bits which means values in the range 0 to 4095.  If you combine 1024 of these and wish to retain perfect precision then you need values in the range 0 to 4095*1024 which is 22-bits.  If your output file is only 12-bits then you are throwing away 10 bits of potentially useful information.  For solar imaging you are not going to notice this because you (probably) don't need more than 12-bits for an excellent solar image.  But for deep sky work where you don't want to oversaturate the main stars of the Pleiades stars but you also want the background dust to appear then you really don't want to be throwing away potential dynamic range by reducing bit depth.  Another good example is a (linear) HDR image of the Orion Nebula.

For some of the images I do, a stacked linear data file of only 16-bits would be far too constraining.

Mark