I am aware that on a 16 bit image, live selections (think marching ants) once adjusted are immediately rendered 8 bit (and now quite prone to posterization issues). Live selections, unless turned into a mask are very destructive, when adjusted.
Having said that, it is my suspicion that by using quick mask to make various selections, to delete (using CONTENT AWARE DELETE) will also render the live selection areas 8 bit.
Is that true?
So are you saying that once I delete, the area(s) will or will not be 8 bit and prone to future posteration issues? I'm not sure what: "the selection mask does not affect the precision of the image data" means. Thank you.
The selection mask is 8 bits, regardless of the document precision.
That fact has no effect on the precision of the image.
No, it won't change the image or cause future posterization -- the image data is still the same precision as it always was.
A mask in a 16 bit image is 16-bit. A selection made from a mask in a 16 bit image is 8 bit. So depending on what you want to do, it's technically better to erase something via a layer mask vs an 8-bit selection and a delete. But the odds that you could EVER detect the differences is very slight. Where 8-bit selections can be problimatic is where you alter or manipulate the actual selection. The engineers looked at updating 16-bit image selections to 8-bit but the need to do so was never really proven–if you can prove the need, I'm sure Chris will look into the issue.
As a fine art landscape photographer making fine art anlargements I was running into posterization problems even with 16 bit images and did not know why. I came here years ago and asked, and Jeff you cleared it up. I was doing a lot of masking but was doing a lot of adjustments on live selections. As an example I would maybe use color range to make a selection then refine edge to feather it further and then I was adjusting that selection. What I learned was, I was damaging my image in those areas (rendering the areas down to 8 bit). I changed my workflow immediately to turning all those types of live selections into a masks (via create a mask on the Layers Pallet). This has resolved the issues. I am a mask hound. Then as I have been critiquing my workflow, lately, I realized I was (sometimes) still doing a lot of live selection work in the area of content aware delete (I use quick mask to make the selections). Sometimes in skies, to remove dust specks and just various distracting elements. I do this early in workflow as I import into PS. Then I got to thinking, maybe I am running the risk of more posterization issues in those areas because they are live selections. Given the fact that my images may still go through a whole bunch more adjustments in PS including a color space jump to LAB and back, I am just wondering if I indeed am potentially creating damage in these areas that could show up in a print.
You were not rendering the area to 8 bit.
You were applying a 16 bit adjustment to a 16 bit image though an 8 bit selection mask with a lot of feathering. Yes, if you did a lot of those, you might see quantization.
Without feathering, it wouldn't make a difference. Without doing a lot of adjustments, it wouldn't make a visible difference.
Honestly, it's not worth worrying about, ACRFREAK.
The likelihood of your manipulating partially selected pixels enough times and in ways that matter so as to make a visible image degradation is tiny. I understand implicitly what's going on (having the benefit of being a software developer of Photoshop plug-ins) and I never worry about 8 bit selections vs. 16 bit masks. Not that I don't use masks, but I use the selection tools as needed, and things just work out fine.
Noel, as a fine art printer charging a lot for my work or doing the work of others it is something to worry about to a certain degree. I have found a variety of way to control tones/colors via masking or blend if sliders or whatever and the build up of a bunch of adjustments has sometimes created damage noticable in the prints. So, I am just trying to figure out how to minimize this by knowing a little bit more about what is going on behind the scenes.
Chris, yes, that does work and is a bit elementary. I'm trying to do a little more advanced ways of choosing very localized areas and adjusting them and keeping the build up of damage to a minimum. Aparently blurring a selection then converting it into a mask is somewhat risky.
First let me say I am not trying to get you to lower standards or accept "good enough". Lord knows, I don't.
And I absolutely support your wanting to understand how things work in order to try to use the tools as well as possible. Bravo for that!
But it occurs to me that if you're using things like Content Aware Fill that replace pixels, the "damage" is may be coming from other places.
Would you be willing to be more specific about a set of adjustments you've done leading to "damage" visible in prints? Perhaps some workflow discussion is in order.
"Creating the mask in 16 bit is a better idea."
Sorry to sound dumb here, but in the above scenerio (color range and refine edge) is that posible?
Side note, I don't work that way much any more.
Yes, add an empty mask to the image layer, target the mask and click "Color Range..." in the mask's properties to access the color range "tool". The resulting mask will have finer gradation than when you do "Select > Color Range..." then convert the selection to a mask.
Remember that the limited bit-depth of your monitor will result in an apparently similar banding in the 16-bit masks created both ways, but analysing the mask pixel values will reveal the finer gradation when an intermediate selection stage is avoided.
Noel, thank you. As I mention above, I am not getting the damage any more. I just wanted to know about potential damage of the content aware delete (and masking live selections). My workflow is erratic at best depending on the needs of an image...
Conroy, THANK YOU that is what I needed to know in respect to masking. I love how on these forums we can access such info!
Europe, Middle East and Africa