Any real life benefits always process 16-bit images in 32-bit Document?

Here everybody can post his problems with PhotoLine
User avatar
shijan
Mitglied
Posts: 1552
Joined: Mon 23 Dec 2019 15:21
Location: Ukraine

Any real life benefits always process 16-bit images in 32-bit Document?

Post by shijan »

This is probably a question directly to developers. Beside trick with lossless negative values recovery in Output Levels, clipless AppleCMS transformations to Lab and back with preserved negative values, is there any real life benefit to process things in "upscaled" 32-bit depth document, then Flatten Image and switch back to 16-bit?

For example, i work with Linear gamma images (film scans and images from microscope), i do Mean stacking of 10 Layers recorded as Action (Load all the individual photos as layers. Keep the bottom layer at 100% opacity. Reduce the layer above it to 1/2 opacity (50%). Go to 1/3 (33.3%) opacity for the next layer up. Then 1/4 (25%), 1/5 (20%), 1/6 (16.6%), 1/7 (14.2%), 1/8 (12.5%), 1/9 (11.1%) and so on...)
Also i do Flat Field Correction with Divide Layer blending mode. as it was described https://www.pl32.com/forum3/viewtopic.php?f=7&t=6660 All remains in Linear gamma.

If i switch document from 16-bit to 32-bit, i can see some tiny changes in Histogram. Is it just a change in visual representation of or Histogram, or is it something really became calculated "better"? I also noticed that in some situations Levels Auto Correction detected a little bit different in 32-bit.

P.S. I remember, it was noticed that Stitching in PhotoLine always works in 32-bit, so maybe upscale to 32-bit makes sense for common workflow?
PhotoLine UI Icons Customization Project: https://www.pl32.com/forum3/viewtopic.php?f=3&t=6302
Martin Huber
Entwickler
Entwickler
Posts: 4183
Joined: Tue 19 Nov 2002 15:49

Re: Any real life benefits always process 16-bit images in 32-bit Document?

Post by Martin Huber »

First the clear disadvantages of 32 bit:
- It needs more memory.
- the resulting file sizes are usually much larger.

32 bit has advantages when your layer calculations are complex and "intermediate" results may be smaller than 0 or larger than 1. 16 bit can't do that.
In addition, the precision of 32 bit is higher.

Whether 16 bits are faster or slower than 32 bits depends on the function and/or processor/RAM speed.

All in all, most of the time 16 bit is sufficient and the better bit depth.