Flatten/Merge to New Layer Bug in CS3?

R
Posted By
rsinmadrid
Dec 15, 2007
Views
385
Replies
9
Status
Closed
I stumbled on this in the course of editing a more complicated pic. The simplest case is as follows:

Open a pic
Duplicate the background layer
Add noise to the copy layer
Change the blending mode of the copy layer
Flatten

In all but a few of the blending modes (normal, multiply, screen, overlay, exclusion), the appearance changes after flattening. It looks as though the blending mode is changing or getting lost. The same thing occurs if instead of flattening, you use ctrl-shift-alt-E to stamp up to a new layer. Using ctrl-E on the copy layer to merge down will work correctly if there are only these two layers, but will fail in the same way if there is a third, hidden layer above the copy layer.

FWIW, I am running CS3 10.0.1 Standard on WinXP Pro, SP2.

If this is not a bug, I wish someone could explain to me what is happening. My mental model of PS says that what I see is what I should get when flattening such a simple case, but that’s not happening here. I don’t get it.

Thanks,
Richard

MacBook Pro 16” Mockups 🔥

– in 4 materials (clay versions included)

– 12 scenes

– 48 MacBook Pro 16″ mockups

– 6000 x 4500 px

DM
dave_milbut
Dec 15, 2007
are you viewing at 100% zoom before and after flattening?
R
rsinmadrid
Dec 15, 2007
No, my screen is far too small to work at 100%.

I just did a quick test on one of the modes, and I’m glad to say, it appears to work correctly at 100%.

OK, so why doesn’t it look right at, say, 25%? Which is correct in the reduced size view, before or after flattening? How should I be dealing with this, given that I can’t afford a huge monitor?

Thanks,
-R.
DM
dave_milbut
Dec 16, 2007
perhaps others can chime in on their ideas of how you should handle it, i don’t usually work that big where it bothers me (sometimes, but not usually) but i do know that anything else than 100% is an interpolation, that is, not an accurate view. to guarantee wysiwyg, you need to work in 100%, or at the least check 100% zoom before flattening.
QP
Q_Photo
Dec 16, 2007
Not sure that I’m correct but it may make a difference as to which layer you have selected when you flatten. I try to always select the background and then flatten. And Dave is correct about 100% being the only accurate view. I work at different zoom percentages but when using filters and so-forth I always check the effect at 100%.

Q
R
rsinmadrid
Dec 16, 2007
Thanks for the helpful information, Dave and Q. I did a little more testing and I am satisfied that what I originally observed was an artifact of the viewing size. On my system, the spurious differences kick in only when viewing at 50% or less. At 51%, everything is fine.

Nevertheless, I still think something is wrong here. I understand that interpolation and compression inevitably introduce distortions. However, I don’t see why two 33% previews should be inconsistent with each other. Obviously, different algorithms are being applied (or different data sets are being used), but it leaves the user wondering which of the two is the better representation of the full-sized image. I tried to compare the two small versions of my test image to the 100% view and I confess that I was unable to decide which was closer. Perhaps that means that it doesn’t really matter, but I am still uneasy.
RK
Rob_Keijzer
Dec 16, 2007
The problem is that a lot of the noise particles are 1 pixel in size. This means the below a viewing of 50% a decision must be made by PS to whether map the particle as one, or as zero pixels, for the viewing.

This is just a viewing artefact. The image’s actual merging isn’t different at different viewing percentages.

I’d say to avoid this, always view (part of) the image as 100%

Rob
P
Phosphor
Dec 16, 2007
I could be wrong, but I think it’s because at percentages less than 100% the image pixels to not correspond to the monitor pixels 1to 1, hence it does not display properly.

It should not matter what layer you have selected when you flatten. Flatten is flatten.
R
rsinmadrid
Dec 16, 2007
Yes, Rob, I think this is a key point. The problem manifests itself when noise is added. In another test with curve and saturation adjustment layers, there seemed to be no observable difference before and after flattening regardless of viewing size, except for hard mix mode, which showed only minor changes.

-R.
R
rsinmadrid
Dec 16, 2007
I don’t see how you can ever have a one-to-one correspondence between image and monitor pixels for a whole pic. Even my modest Canon 20D produces twice as many pixels as an Apple 30" Cinema Display offers at maximum resolution. So there is always going to be some compression and there may be differences (hopefully minor) between the screen image and a printed version. But that is a different problem than having inconsistent preview images on the monitor.

-R.

Master Retouching Hair

Learn how to rescue details, remove flyaways, add volume, and enhance the definition of hair in any photo. We break down every tool and technique in Photoshop to get picture-perfect hair, every time.

Related Discussion Topics

Nice and short text about related topics in discussion sections