This content has been marked as final. Show 30 replies
Lens tab and Detail tab:
On reflection, the first point is pretty obvious (I think!), but it is probably worth saying anyway.
> Get the chromatic aberration and defringing sorted out in the Lens tab before trying to sort out sharpening in the Detail tab.
My current order in the Lens tab is (at 400%): address Red/Cyan fringe with the Alt key held down; then sort out Blue/Yellow (ditto); then re-do the Red/Cyan.
Then try to defringe if necessary. So far, I haven't found a case where the optimum setting is the middle "Highlight edges" value rather than one of the others, but perhaps I'll find some.
In fact, nowadays I try to sort out the CA/defringe in the Lens tab before just about anything else in other tabs. Is there any case where other things need to be sorted out first? It gives me a cleaner image to work with elsewhere.
Detail tab, especially Sharpening (fur & feathers):
I can't find an optimum order to sort things out in this tab. Each slider appears to affect the results of each other. So what I am having to do is make some initial guesses based on the subject, then tweak things from there.
About the one thing that in the past I have often found ACR sharpening good for (not necessarily better than Photoshop, but good enough for capture sharpening) is "in-focus fur and feathers". So I've just tried some such cases where ACR worked OK before. I believe this is a case where ACR 4.1 is even better. But the slider values I'm getting are a bit extreme.
(I'm using Jeff's tip of trying to get it right on screen at 100%, not crunchy).
Only increase the Luminance slider if there are also significant areas that are not fur and feathers. Avoid harming their texture. (On occasion, I've done 2 conversions, one for the fur & feathers and one for the rest, with different sharpening and noise sliders, then masked the layers appropriately in Photoshop).
Masking - 0. I'm trying to emphasise details/texture as well, not just true edges. (And I've found that if I can get the detail of fur and feathers right, noise is less important).
Radius - 0.5. (It depends on how fine the fur and feathers are, but start low).
Detail - 100! Perhaps it will be necessary to back off a bit, but with low values for Masking and Radius, this starts to address the fine detail.
Amount - adjust to taste, perhaps 30 or more. With Detail so high, it probably won't be a very high value.
I've just had a horse's hair looking ready to stroke on the screen! And every strand of an owl's feathers round the eyes. But apparently without nasty artifacts elsewhere.
(I expect that, over time, I will settle on less extreme values. But it is worth trying them to see what happens).
(I've already said this elsewhere, but I'll repeat it here).
A comment about "Clarity":
I've just opened some earlier images that had the problems of flare from specular highlights. A bit unexpected, to me, was the improvement I got using a high value of Clarity. The effect was to get the lost contrast as a result of the flare under control - it was as though the flare was being reduced, and features partly "masked" by the flare became stronger.
Perhaps this was expected by others - but I thought Clarity was going to be most useful on mid-tones rather these lighter areas. It is worth a bit of experimenting. In one case I had previously resorted to correcting the problem in Photoshop, but Clarity would have reduced (perhaps not eliminated) the need for that.
I can imagine that it would sometimes be useful to do 2 conversions, one with LOTS of "Clarity" to reduce the problem of flare, and another to suit the rest of the image, and use masking of 2 layers in Photoshop as appropriate.
Details tab, especially sharpening ("man with character"):
Don't try this with a woman unless you want to be ostracised!
I've just been re-processing a couple of portraits of men who have "character" in their faces. Somewhat weathered skin, a bit of stubble.
The best settings I've found were the same as #2 (fur & feathers):
Masking - 0.
Radius - 0.5.
Details - 100.
Amount - to taste.
I've just been re-reading Bruce's "Image Sharpening" book - page 167. He mapped USM sharpening radius to camera megapixels, for capture sharpening, as (summary): 6 MP - 0.6 pixel; 8 - 0.5 pixel. (Etc).
I suspect things aren't as simple as an exact correspondence between those values and the Radius values in the Detail tab! But ... with my 6 MP *istD and my 10 MP K10D, all the results so far are better with Radius 0.5 pixel than the default 1.0 pixel. (When viewed on the screen at 100%). I intend to set this as my new default.
Good info Barry. I had to step in to save you from a protracted soliloquy :)
> "I had to step in to save you from a protracted soliloquy"
I find that one of the best ways to see if I understand something is to write it down then read it back as though someone else is reading it. In each case above, just writing it down clarified my own views, and sometimes changed them.
I think a pattern is emerging in my tests, and not surprisingly they relate to what Bruce Fraser said in his book about "Optimising for Image Content", with different parameters for high-frequency and low-frequency images.
(I'm currently playing with some low-frequency noisy images, to see when to use Luminance Smoothing and when to use Masking, or combinations. They do different things, but sometimes with an image that has some noise it may be best simply to avoid sharpening the noise rather than necessarily to reduce it).
It occurs to me that perhaps I should wait for Lightroom 1.1, then use Snapshots to compare different combinations of these sliders. I am gradually moving away from ACR 4.1 towards Lightroom.
> It occurs to me that perhaps I should wait for Lightroom 1.1, then use Snapshots to compare different combinations of these sliders. I am gradually moving away from ACR 4.1 towards Lightroom.
I need to probably follow your posts in this regard. I too am going back and forth. At this point LR seems to be not quite up to speed so I do most of my proecessing in Bridge/ACR CS3 but then go to Lightroom for the galleries.
I do like Lightroom but am waiting for the update before using it for actual processing. Comments and observations like yours are helpful since I don't have as much time to experiment.
Lightroom 1.1 will have the same basic processing engine as ACR 4.1, of course, and, hopefully other significant improvements. In terms of basic processing parameters, of course, there will be no significant difference that I can see, so any experimenting with new features in 4.1 will be applicable to LR 1.1.
However,I am presently using Bridge 2.1/ACR 4.1 and IdImager as a DAM. My timed workflow doing the exact same processes on a batch of 100 mixed format images from ingestion to db is 60% of what it is in LR--at this time, that is.
Hopefully LR 1.1 will be faster, but I somehow doubt it. This 1.1 upgrade is likely to be a 'fix what's there and add a few features' effort to get it stable and fully functional. Later upgrades are where I would expect to see serious changes and improvements to Library, which is the present bottleneck in LR and what keeps it from being a comptetitive production tool.
Are the Galleries in Lightroom any different from the ones that you can now add to Bridge CS3 v.2.1 by installing the new AMG (Adobe Media Gallery) software?
I find those AMG galleries very quick and easy to use because you can create a Gallery, and upload it directly to your website, from within Bridge.
You can download the AMG software from here:
The instructions make it sound more difficult to use than it actually is.
Just read a little and then just experiment and you will soon get the hang of it.
I just cobbled this Flash-based Gallery together very quickly in Bridge from a few of my existing JPEGs:
Thank you for the links Ann. I think I started on this the other day and got pulled away before I could finish, then forgot about it. I love your gallery. And my favorite photo of yours is first (Thames, right?). I still find that photo fascinating.
Thank you Linda. And it was indeed taken on the Thames.
Up to now I've been posting views on working with high-frequency images, such as "in-focus fur and feathers". I've recently been trying to identify generic parameters for other cases. This is harder.
Working with images with a bit less high-frequency detail, and some noise (ISO 400 with a Pentax K10D), I have come to the following conclusion: it is more important to avoid sharpening noise than it is to reduce noise before sharpening.
This means that the Masking slider is more important than the Luminance noise-reduction slider. I find that I don't like the look of high levels of Luminance noise-reduction - it can give a weird "posterisation" effect (I don't know how to describe it).
This may be related to a technique that I have sometimes used in Photoshop to reduce posterisation - I inject a bit of noise into the relevant areas, either in the image or in a layer mask. The noise is more acceptable than the posterisation.
So, for these images, I am staying with default Colour noise-reduction (25), and minimum Radius of 0.5, and only increasing Luminance noise-reduction as long as there is no hint of any weirdness. Perhaps no more than 25, typically.
The Details slider is adjusted according to whether it is still a high-frequency image, so may still be quite high. With an Amount that may be temporarily exaggerated, I increase the Masking until the noise is acceptable. (It is unlikely to be perfect). Perhaps even 50. Then the Amount is dropped until there are no adverse affects visible, certainly at 100%, preferably at higher magnifications.
So far, when trying to do capture sharpening in ACR 4.1, I'm ending up with a Luminance noise-reduction value less than the Masking value. I'm sure it isn't as simple as that, but I'll continue like that for a while until I learn something better. And ... I am very comfortable with a Radius of 0.5 for both my 6MP and my 10MP camera.
Thank you for keeping us informed on your observations. It's helpful to see how you test and think things through.
The more experiments I do, the more I revise some of my previous views! Oh, heck!
I've just been processing images taken at ISO 200 (K10D) - a little bit of noise visible at 100%, but not objectionable.
I've found there is a case for using Luminance noise-reduction even with zero-Masking, contrary to #12. But only small amounts, eg. 10. I REALLY don't lke the look of heavy use of Luminance noise-reduction. (I never have, and it hasn't changed much with ACR 4.1).
So, at ISO 100 with the K10D, or ISO 200 with the *istD, I think Luminance noise-reduction of 0 is likely to be a good default. Perhaps rising to about 10 when ISO is a step higher. (But for a high frequency image, such as in-focus fur and feathers, 0 in that case too). Then rising to perhaps 25 for noisier images.
(If it is noisier than that, convert to monochrome and pretend it is grainy B&W film!)
A cautionary tale about "Defringe".
I had some very bright spots that for some reason I couldn't fix completely with the Red/Cyan and Blue/Yellow sliders in the Lens tab.
I could finish the job using Defringe. But that affected other highlight areas, surrounding them with an unpleasant desaturated grey band. (This resembles the effect of totally blown highlights being handled by a futile excessive use of "Recover" - yeuk!) This was worse than the problem of CA.
Lesson - Defringe is not there to fine-tune CA-reduction, and can have a bad effect elsewhere. Sort out the CA, then live with the results. Use Defringe for genuine purple fringing or similar, after CA correction, but watch out for unwanted consequences.
Which basically goes back to my original article on PSN...it was Thomas' desire to improve both the demosiacing and noise reduction for CR 4.1. He did. Now, some people would like to have Camera Raw have the heavy duty noise reduction of say, Noiseware...but that's really not Camera Raw's job. Camera Raw is targeted at doing only those things that are BEST done during the raw-color space converted image. Some things (like heavy duty noise reduction) are better done after raw conversion where you have the ability to create localized corrections and layer blendings...
A well exposed ISO image prolly doesn't need any luminance noise reduction...true. But an ISO 200 (or under exposed ISO 100-which is basically what an ISO 200 image is) WILL benefit from a "touch" of Camera Raw's noise reduction-the amount is up to you to decide. If I do a .3 Exposure bump to an ISO 100 image, I would look at doing a 5-15 luminance noise reduction to help reduct the additional noise created by the Exposure increase (caused not by Camera Raw but the additional noise in the shadows which are being forced up)
All this stuff will vary greatly by camera, by ISO and by Exposure settings...and while CR 4.1's noise reduction is clearly better, it doesn't take the place of (in fact doesn't even try) heavy duty post-processing software like Noiseware...
I found the unpleasant gray band even before defringe was avilable. It was exacerbated by further adjustments, to the extent I had to go to the pixel level and remove the most obvious examples.
Gotta avoid the more knee-jerk, obsessive moves, IMO.
Jeff Schewe wrote:
>Which basically goes back to my original article on PSN...it was Thomas' desire to improve both the demosiacing and noise reduction for CR 4.1. He did. Now, some people would like to have Camera Raw have the heavy duty noise reduction of say, Noiseware...but that's really not Camera Raw's job. Camera Raw is targeted at doing only those things that are BEST done during the raw-color space converted image. Some things (like heavy duty noise reduction) are better done after raw conversion where you have the ability to create localized corrections and layer blendings...
>All this stuff will vary greatly by camera, by ISO and by Exposure settings...and while CR 4.1's noise reduction is clearly better, it doesn't take the place of (in fact doesn't even try) heavy duty post-processing software like Noiseware...
While I would agree that heavy duty noise reduction is best be done in PP, this would mean that all sharpening should be disabled in ACR so that the noise is not sharpened. We are then back to pre-ACR 4.1. As Bruce points out in his sharpening book, noise reduction and sharpening give conflicting effects, and it is necessary to achieve a proper balance. IMHO it would be best to combine them in a single software package. Noiseware and other NR programs have sharpening components, but I don't know how smart they are.
ACR 4.1 already has an edge mask for its sharpening and it also incorporates the functionality of Bruce's layer blend if. It would not seem that much more to add a surface mask for noise reduction to protect the edges. While one might get more control of the NR in PP, the various options in Noiseware do give quite a bit of control and it is conceivable that one could get reasonably good sharpening and noise reduction if such functionality were built into the raw converter. As you say, noise is not a problem with properly exposed images at low ISO, it can be a significant problem at high ISO often needed for event and sprots photography. BibblePro already incorporates Noise Ninja. Finally, one could also argue that heavy duty sharpening could also best be achieved in PP; for example, one might mask out clear blue sky so that it would not be sharpened.
I have observed that, for high ISO shots (Canon 10D), setting (still) just the Color Noise setting at a moderate setting, combined with a masked, low sharpening in CR 4.1, is a great starting point for further denoising (Neat Image, Noise Ninja) and sharpening (Smart Sharpening). I'd much rather have a deconvolution sharpener in CR though, for various reasons. For instance, I cant control the impact of the halos precisely enough (gotten used to Smart Sharpens sliders, you see).
Edit: I find the concept of masking the single most important lesson in Photoshop (AKA Chops) - this in regards to denoising as well as sharpening. It can make all the difference if you find the right type of edge mask. Some further options in CR in denoise (coupled with sharpening mask) would be most welcome.
Mathias, have you tried the Reindeer Graphics edge finder?
Not extensively. I might have tried it once, but lost the filter in an update. I'll look into it, thanks. I do have my own set of edge finder actions, that serve me well.
Are they the same as ones from several years ago? I had some of your actions and lost them.
I dont know... My favourite edge finder action is called the Roberts algorithm, and is a custom filter that 'lights the edges' from 8 different angles to bring out all the little details.
Reindeer Graphics edge finder has anybody gotten it to work in PS CS3?
I have. No problems, AFAIK.
Why would someone want to use the Reindeer Graphics edge finder when Photoshop already incorporates an edge-finding filter? Is there something wrong with Photoshop's built-in filter?
Try it and see for yourself. It's free, and has been around when PS edge finder was nowhere near as good.
Bruce's book was what really opened my eyes to the power of ACR. Before I read his book, my attempts at RAW conversion were unsuccessful. I usually ran back to my old trusty 'levels and curves' workflows in photoshop.
I just installed CS3 yesterday and saw the new sliders...first thing I thought was "I will miss Bruce's book"
Are there good tutorials on the new sliders? Haven't had time to look but I am a new member of NAPP.
See Jeff Schewe's comprehensive article on what is new in acr 4.1:
Also, Jeff will be updating "Real World Camera Raw" in a few months.