32 bit vs 16bit, HDR merge

[Jan Barta]

Hello,

I've just finished watching all interesting tutorials from affinity (to me) and I very like the 32 bit editing. And I am wondering what is the drawback while using the 32 bit? The export says the same file size with 16 a 32, but a user gets much better file to work with.

The biggest advantage I see in keeping the colour/grade information even after over/underexposing the file in develop person (using layers keep it in both - PS don't) and then, you can modify the exposure again with no pixcel losage.

Also, would it be posible to make develop persone work in two settings.
1. as it is now - all edits are destructible to the image.
2. live settings - all edits are remembered (for example detail refinement, noise reduction and white balance is something a user would like to change during editing. I know white balance can be modifed, but a user loses the Kelvin scale).

My standard workflow is using lightroom (non destructive edit) to keep the same colour grading to a series of photos and for harder editing using the PS or now AP. At some point or next day I don't like the colour and I can easy modify the settings.

In some comments I read, if using apply tone curve, the file loses dynamic range - could you clarify this? What does this actually do to the image?

The detail refinement option is set to zero at default. At first I was shocked why a file loaded into AP is so blurry compared to viewing in basic picture browser (faststone viewer), then I found the DR option.

During HDR merge could the noise reduction option be at develop persona, so that a user sees the effect and can adjust to his liking (detail can be done so)? 

Thanks, Jan.

[NotMyFault]

Hi Jan,

 

your post has many questions and suggestions, please let me allow to focus on the first of them:

• And I am wondering what is the drawback while using the 32 bit? The export says the same file size with 16 a 32, but a user gets much better file to work with.

1. Performance may degrade while editing in Photo. E.g. try a curves adjustment with LAB color format.
2. You need much more RAM for files with high pixel dimensions
3. some functions are unavailable in RGB32, e.g. frequency separation, dust & scratches filter,
4. When you use color gradients (e.g. for vector shapes, pixel layers), and "merge visible", photo applies (unwanted) dithering
5. The photo files are much larger (before export)
6. In RGB32, you use a linear color profile - this could lead to unexpected results when converting and exporting

RGB32 is quite valuable for very specific task like HDR merge. But i often get files > 1 Gbyte from simple CR2 raw files of ~25 Mbyte after a few edit steps.

When you do not need these extra color bit depth, RGB/16 or LAB are more practical in every aspect.

Regards, Timo

 

[Fixx]

11 hours ago, Jan Barta said:

In some comments I read, if using apply tone curve, the file loses dynamic range - could you clarify this? What does this actually do to the image?

It is due rounding errors. When curve tool assigns new values to pixels they are approximated to nearest pixel brightness level. Normal RGB has only 256 of those so there is not too much to choose from.

Maybe "dynamic range" is not the right term as it relates to ends of range (blacks and whites) and usually curve adjusts mid pixels (though if you manipulate end of the curve it certainly affects range ends).

Severe problems show as banding as there are less colours available. That said, even in normal 8-bit / channel RGB edits should be biggish this to show and with 16-bit images there should be nothing to worry about.