Content Aware Fill at 16-bit? (i.e. without active selection)

[Ilario5E19]

Hi guys, newbie here. Quetsion: is it possible to apply Content Aware Fill at 16-bit? I mean, I knew that any active selection reduces the bit-depth to 8-bit. So, is there a way to use Content Aware Fill without selecting an active area? Thanks in advance.

[Mohammad.Harb]

Content aware fill requires selection . It cannot be done without selction .

[TheDigitalDog]

---

@Ilario5E19 wrote:

 I mean, I knew that any active selection reduces the bit-depth to 8-bit. 

---

The data is the data; so if you have high bit data (Photoshop calls anything more than 8-bits per color "16-bits" even though it's not), the content aware will be high bit. The selection is moot in terms of the data being processed. 

As Mohammad stated, you must make a selection so by all means do so. 

[Trevor.Dennis]

Did you know that pixel-peeping can drive you mad?   Photoshop's automatic fixes have evolved over the years and are truly amazing, but they are a long way short of perfect, and likely to leave compromised results far in excess any quality loss from a temporary localised drop to 8 bit.  That's if that even happens.

If you are outputting to a very large high quality print, then it might pay to obsess about IQ, but if that was the case then COF is not the best tool for the job anyway.  If you have a subtle gradient that covers a significant percentage of your image, then 16 bit will be advantagous.  Same thing with greyscale/B&W.  The bottom line is can you see a difference, but a lot of the time a computer monitor is not the best tool for discerning that difference.

Our @D Fosse  does museum quality work, and has to maximise quality, so I wonder what he would say?

[D Fosse]

This isn't a big deal. Although it's correct that selections are 8 bit depth (legacy code), the underlying data are still 16 bit as Andrew points out.

If you use the selection to make a mask, the mask itself is still 16 bit, but the data contained in that mask will have 8 bit stepping. So you can in principle get into banding trouble if you make extreme adjustments based on that mask/selection. In practice - meh, not a big concern.

If you feel particularly paranoid that day, you can get around it by copying Lab L data and pasting it into a new mask. That's a pretty complex operation by itself, so it pays to make an action to do it if you feel you need it. I have one, but rarely bother to use it.

To be clear - painting directly into a mask is entirely end-to-end 16 bit and not affected.

[Ilario5E19]

Thanks so much for all your advices guys!

@D Fosse Many thanks for your kind reply. Just 2 questions:
1. Just to be sure, when I scale with Content-Aware Scale (or with free Transform), I'm still working at 16-bit, right?
2. Can you kindly elaborate the Lab L method? It sounds to me like a Luminosity mask method... Sure! I could make a black “hole” with the paint tool directly on the “wrong” area of the image, then create a mask for black-only with Lumi32 (that I own) then choose Select! This produces an active selection, it's true, but this time being made by Lumi32, the selection should be at 32-bit (in spite of being an active selection) because not made with Photoshop code. Does it make sense?

[D Fosse]

Content-aware fill works best with hard selections. So the whole question is moot in this case and doesn't apply - there won't be any soft gradations that could cause banding issues.

Yes, the Lab method is for luminosity masking. Basically, you convert to Lab, target the L channel and copy all. Then go back to RGB (undo will work), and paste into the required mask. It's not really complicated as such, there's just a number of steps to remember, so an action makes it much easier.

But again - if you're making this mask by painting, this is not an issue at all. It will stay 16 bit the whole time. The 8 bit limitation is only for selections.

[TheDigitalDog]

---

@Ilario5E19 wrote:

Thanks so much for all your advices guys!

@D Fosse Many thanks for your kind reply. Just 2 questions:
1. Just to be sure, when I scale with Content-Aware Scale (or with free Transform), I'm still working at 16-bit, right?
2. Can you kindly elaborate the Lab L method? It sounds to me like a Luminosity mask method... 

---

1. Yes. 

2. Sorry: Pet Peeve concerning the misuse of the term Luminance (which has been in Photoshop and other applications for years but doesn’t make it a correct term). 

Luminosity is a measure of the total radiant energy from a body. It has nothing to do with what a human observer perceives but rather describes the total radiant energy, such as watts/second of a source (the surface of a radiating object like a display). In Photoshop, the layer mode called luminosity is not what's happening here (I was told its something like the "Luma" which is an old TV RGB transform)

Lightness is a perceptually scaled component of color, the axis seen in Lab (Lstar) from light to dark. It IS the L in HSL. 

[D Fosse]

Re-reading my own post, I don't think I was perfectly crystal clear. So to remove all doubt:

Even if the selection is 8 bit, the image data contained within that selection is still 16 bit.

It only becomes an issue if you use that selection for extreme tonal adjustments and the selection has smooth gradients.. Normally it's a non-issue.