If the circumstances force me to work on images with noise which have to be enlarged, and if I use noise reduction based on noise-pattern-recognition, is it better to carry out the noise reduction before or after the enlargement (and in case the answer is after, which resampling method would resample the image (and thus the noise) best for subsequent noise reduction?)
And what's the logical rationale behind the answer ?
Thanks a lot!
PS: I think the answer will be the same for any type of noise, but if there are significant differences between noise types, I'd be interested in that, too.
And besides, even if I had seen from experience which way has better results, I guess I'd still have posted to understand the reason for it.
Bill
Rob
Another useful trick is to do noise reduction inside a reverse luminosity mask, leaving the highlights more or less unaffected.
Another useful trick is to do noise reduction inside an inverse luminosity
mask, leaving the highlights more or less unaffected.
Can you explain this Freeagent?
I also thought that NR should be applied as soon as possible, leaving the noise patterns intact. On the other hand I thought that enlarging also enlarges residual noise, so maybe the bottom line is to do a regular NR initially and then, after enlarging, a second weak/careful one on parts of the image if need be.
@Freeagent: very good tip the inverse luminosity mask! As for ACR: From my personal experience, ACR noise reduction is very effective on RAW material, but on all other sources it far lags behind noise-pattern-recognition methods.
Hey! Is this telepathy?
Yes, I figured it would be simpler to just beam it over ;-)
Anyway, if that was unclear to anyone else, it means ctrl-clicking on the RGB channel (or Luminosity in Lab for that matter) to load it as a selection, invert it so the dark areas are more selected, and then noise reduce (reduct?).
The point is that noise is a bigger problem in the shadow/less exposed areas, while the highlights may in fact be OK. And that's where most of the important detail is.
I've found ACR to do well on TIFFs and JPEGs as well, but of course there's no formula. The main advantage is to combine noise reduction with the "0-detail-slider" sharpening (and perhaps mask) to keep as much detail as possible while still getting rid of most of the noise. And you can check the combination in real-time while you play with the sliders.
You know how it is...you try to explain but the words get in the way...