The first step, always, in troubleshooting this anomaly is to prove that it is really happening. Make sure you zoom to 100% – any other maginification is an approximation of the image. Then, flatten your image – use the history palette to toggle "before/after". 99% of the time, it’s only a visual anomaly, not a real artifact.
Peace,
Tony
Tony is right.
The other 1% of the time it can be because of the way certain blending modes react upon flattening. If it turns out it is and issue, experiment with "Merge Down" and "Merge Visible" as well as "Flatten"
Chris
I am in no way absolutely sure about this but; I think it may make a difference as to which layer you have selected while merging or flattening. I find that the result may be different if I have selected an adjustment layer and then flattened. Results seem more accurate if I highlight the background layer before I flatten.
Peace,
Q
Q ,
I’ve observed the same phenomenon on occasion and wondered whether I was crazy or not. Good to have a companion in craziness .
π
Brent,
We have established that we both have experienced this phenomenon. Sadly, it does not rule out that one, or possibly both of us, is/are still crazyΒ
Q
I would like to see somedody who sees this type of thing flatten a few different ways, by flattening with different layers highlighted, and by merging or stamping down.
Then compare all the flattened images via the apply image command.
I would wager there is no difference.
Q,
We probably share a buggy video driver ! π
If it exists ( like the Loch Ness monster ) it’s a pretty elusive critter. Here’s an idea ( like the Loch Ness monster ) , I’ll offer a $20 reward to the person who furnishes undeniable and repeatable proof of it’s existence. That ought to get everybody working ! <G>
π
Actually, I have a file in which the colors change dramatically when flattening, one I have shown here a couple of years ago.
Tony is right in that at 100% WYSIWYG. But, in my case, 100% gave no hint as to the real look overall. I wound up getting a mustard yellow switching to a burgundy value when flattened, completely destroying the effect I wanted.
Now, here’s the solution. As I came back from 100%, dropping in percentage increments, I noticed that the color shift became a function of the viewing percentage. Trying to get a better grip on it, and because the size of the file made flattening take forever, I changed file size. Now, when I go from 100% to screen, less color change. Hello! So, I dropped the resolution to 72 dpi and now, 100% color is the same as screen size color, and the screen sized color is what i wanted. So, I began increasing the resolution and found that if the file is at 100dpi or less, 100% is the same color as fit screen. The file flattened with no color shift. Actually, with acceptable color shift at 100dpi.
Since the project was for a rug design, 100dpi is acceptable, as the best rug weaving density with hand knotting seems to be at 100 knots/inch.
How convenient!
I hate those color changes, even the minor ones. There seems to be no way to have it all, however.
FYI, the file started at 1800 dpi b4 flattening, dropping to 100dpi.
Now here’s what is really curious. If 100% is accurate, why did the flattened file maintain the colors at fit screen when the resolution decreased? It appears that the anomalys are maintained, rather than accuracy.
I wondered if it was a platform problem, so i took the file to a G5 and saw the same effect. Moreover, the G5 took longer to flatten than the PC.
It is annoying to be sure but the flattening process does nothing different to the actual file color than the stamping down or merging process that some people tout as a fix for the color shift on flattening.
In your case, where you have a pattern of different colored pixels (a rug weave) you have to expect some downsampling error when the thing is viewed at a lesser percentage.
What do you expect the application to do when wildly-divergently-colored file-pixels get smashed into a single screen pixel?
Something has to give somewhere and the zoom factor has everything to do with what you see.
I feel your pain however. I remember working on cloth pattern scans and it is truly maddening. The downsampled view contains downsampling errors and the 100% view is way too miopic
I’d guess that somewhere you had a pretty severe adjustment layer, and you were viewing the image zoomed out instead of at 100%.
Only a 100% view is really accurate, because subsampled images will blur high frequency detail and give a false impression of the adjustment (or filter in many cases).
It’s a mixed bag, Chris. I wanted the look I obtained,at the screen size, which isn’t instantly apparent at 100% that something was wrong. Now, when I do go to 100% I know what to look for. Then, I only zeroed in around the center to inspect sharpening and look for funny edges or misalighnment.
There were some severe adjustment layers as you surmise. The root image is a b&w shot of basalt. Through mirroring and pattern repeats I got what you see. I was thrilled….until I flattened.
I tried to approach it systematically by flattening as I went along. That got me even further away.
So, now it remains a curiousity in my file. At 100dpi, too much detail is lost as a straight image.
It still puzzles me that the color would be so far different. Detail ok, but that kind of color shift….
Even at 72 dpi, the flattened file is too saturated and the contrast and brighness is off. It needed a bit of de sat and some curves work to match the unflattened look.
Y’all- Aniother facor might be "opacity." If one or more of your layers are at less than %100 opacity in a given image, flattening will certainly(?) cause a change in appearance!
D~
Dennis,
I am curious as to how you draw that conclusion. Using that analogy, if I add an adjustment layer of red with blend mode of multiply, and set that layer opacity at 50%, then when I flatten the color will be… what?
It seems that you are suggesting that we should expect a change if the opacity is less than 100%. I’m sure that’s not the case, but am interested in your rationale.
Peace,
Tony
I think he is confusing the 100% view with 100% opacity.