Sharpening and Noise Reduction
I've been studying COP7, along with Lightroom 5, and while granted I'm pretty new to both these programs, I've noticed i can't get detail up without getting pretty ratty looking sharpening in COP7. I've tried to balance out noise reduction and sharpening, but I'm either getting really ratty sharpening or detail loss from noise reduction.
By contrast, I can get quite a good balance of detail enhancement and noise reduction in Lightroom without much effort at all. I'm under the impression both of these apps use unsharp masking for their sharpening engine. I'm not sure what's the best way to show what I've done in either app, other than to show the final results. It just seems I can crank sharpening up significantly in Lightroom without getting unsharp mask ringing artifacts, whereas they seem to come up rapidly in COP7. Below is a link to a screen grab of a high ISO test file I was using, shot at ISO 1600 (click on it to go to 100%). This screen grab will also show the results of the native camera JPEG, as well as the image processed in DxO 9 Elite, which actually yielded the best detail/noise reduction results, although it's new noise reduction engine is computationally intensive.
So basically, is this the best I can expect to get from COP7 with respect to sharpening/noise reduction? Is there a different sharpening algorithm that COP7 can use other than unsharp masking? Thanks for any input on this.
http://www.mediafire.com/view/22whc9yey ... ompare.jpg
By contrast, I can get quite a good balance of detail enhancement and noise reduction in Lightroom without much effort at all. I'm under the impression both of these apps use unsharp masking for their sharpening engine. I'm not sure what's the best way to show what I've done in either app, other than to show the final results. It just seems I can crank sharpening up significantly in Lightroom without getting unsharp mask ringing artifacts, whereas they seem to come up rapidly in COP7. Below is a link to a screen grab of a high ISO test file I was using, shot at ISO 1600 (click on it to go to 100%). This screen grab will also show the results of the native camera JPEG, as well as the image processed in DxO 9 Elite, which actually yielded the best detail/noise reduction results, although it's new noise reduction engine is computationally intensive.
So basically, is this the best I can expect to get from COP7 with respect to sharpening/noise reduction? Is there a different sharpening algorithm that COP7 can use other than unsharp masking? Thanks for any input on this.
http://www.mediafire.com/view/22whc9yey ... ompare.jpg
0
-
[quote="NNN635334001806491907" wrote:
.... Is there a different sharpening algorithm that COP7 can use other than unsharp masking? ..
A tough subject to deal with, as I find most of the comparisons are going to be subjective.
First, a simple answer to your question - to the best of my knowledge, CO uses only the one sharpening algorithm.
Another comment. I only shoot in raw, and my limited comparisons have told me that LR processes jpegs better than CO; but if I take the original raw image, I can generally do better in CO.
A question - how was the post processing of your OOC jpeg done by the camera wrt sharpening? When I have done noise reduction (NR) post processing in 3rd party packages, I can always obtain better results if I don't sharpen in either LR or CO. Both LR and CO leave artifacts that other NR software can't deal with. I assume that any "in camera" sharpening will also leave some artifacts that will effect NR.
For my higher ISO shots like yours, I tend to get "the best" results when I use CO (or LR), and I send a tiff that does not have any NR or sharpening applied to it over to PS. There I use a Noiseware plugin to eliminate the noise followed by a Topaz plug in called Detail for sharpening. I find the Topaz DeNoise almost as good. Both NR packages are better than LR or CO.0 -
Thanks for your reply. You're merely looking at JPEG outputs from the same RAW file, I was not working with a JPEG source. The in-camera JPEG was the only actual JPEG I displayed. I merely exported from each app in JPEG.
It seemed that no matter how I adjusted the sharpness settings in COP7, I would always wind back up in the same place if I tried to bring up detail the way I did in Lightroom and DxO.
As you can see in the outputs, DxO did the best job overall, albeit at a heavy computational price. It has not only an advanced noise reduction engine, but processes to maintain out-of-focus areas and, more importantly, the transitional focus points from getting "stratified", as you can see in the COP7 and Lightroom outputs.
But that's neither here not there. I'm studying LR and COP7 because they are often requested programs when looking for work (I'm a photographic retoucher by trade). I need to understand and be able to use these apps to their best abilities, hence this post concerning this aspect of image data manipulation. I may have stumbled into a shortcoming of COP7, or there is simply something that needs to be done differently within this app. I can't really imagine what that would be however, as making the sharpening and noise reduction adjustments in both DxO and LR were simple, straightforward processes.0 -
Which camera?
I ask, because I find this to be a key determinant in whether Lr or Capture One is the better converter at high ISO.
Incidentally, Lr uses more than just USM: depending on where the "Detail" slider is positioned (from around 50 onwards), it blends in a Deconvolution algorithm.
Some useful reading here:
http://forums.adobe.com/thread/882952
I have to say that (following a casual observation by a user on here) my favourite "sharpening" tool at the moment is the Structure slider in Capture One's Clarity tool (with the method set to "Punch"). It really beings out microdetail without generating artefacts - I often use this with Capture One's default sharpening dialled down or disabled, and the results are very satisfying.
It's a shame that Capture One doesn't have an equivalent of Lr's Masking slider - that can really make or break an image.
To get back to my original point: with my current camera, I far prefer Capture One's high ISO rendering to that of Lr: it's actually up there with DxO Optics Pro's default (not "PRIME") NR, which is in itself considered to be a class-leader.
This isn't a crop, but it's still pretty bloody good - a file from my Canon 70D at 10,000 ISO, straight out of Capture One, resized and selectively sharpened in PhotoShop. Very little light (only getting 1/40 at f/2.8 here - 10,000 ISO, remember) and no additional NR.0 -
Thanks for you reply. The camera used here was a Nikon D5100. I'll try the structure slider to see if it helps, but last time I used that it was a bit too contrast and saturation inducing when set to punch, I thought.
I would question whether camera model would make any difference, I've dealt with sharpening images in Photoshop with a multitude of cameras and the end result of the different sharpening tools have always been the same. There's nothing uniquely different going on in any of these programs with respect to the algorithms used. Unsharp masking is unsharp masking.
The noise reduction engine in COP7 appears to be OK, it's just that the detail enhancement tools are lacking. While your image appears to look good, it isn't at 100% as you mentioned, and the subject matter doesn't have the same level of fine detail as my example.
I'll try messing around with the settings some more to see if I can improve the output from COP7.0 -
[quote="NNN635334001806491907" wrote:
I would question whether camera model would make any difference
Disagree by all means - but this is several years and many tens of thousands of images-worth of actual experience talking, across a number of cameras and brands, processed in Lr, Capture One and other converters.
And I know how to use these converters.
There's no question that the camera body matters: it's pretty obvious that some cameras just get better noise profiles created for them - human error/lack of care/lack of diligence/lack of comprehensive effort at the profiling stage, I've no doubt - but the evidence speaks for itself.
As an aside, is there a single Real World benefit to be gained from evaluating noise vs. detail at 100+%? Hasn't it been pretty thoroughly demonstrated over the years that pixel peeping simply serves to obscure rather than to illuminate?
You literally can't tell anything useful from how an image looks at the pixel level. I simply do not agree - nor has anyone been able to prove otherwise - that we can only "properly" evaluate noise and performance by looking at 100%, any more than I'd accept an argument that (say) Frans Hals made such a great job of the Laughing Cavalier because of how hard he looked at each individual brush-stroke.
Nope - the detail matters only in the context of the whole painting. If you get in really close, it looks pretty crappy, too.
So unless you're presenting your images at the equivalent of an actual pixels crop, evaluating at the image level is the only meaningful way to compare.0 -
Thanks for your reply. I guess we'll have to agree to disagree there. Images processed from my D600 exhibit the same sharpening anomalies in Capture One as my D5100 images. The processes in both Lightroom and DxO both display better sharpening effects then what I'm getting out of Capture One. And yes, all at 100%. For my own personal work it's not an issue, as I can easily use either LR or DxO as needed. It will be an issue if I will be in a work environment where they will be using Capture One that the sharpening issue will be problematic. Actually, I work in one such studio now and was wondering why they weren't sharpening in Capture One. Although I haven't asked, I'll bet they've run across the same issue and subsequently do all their sharpening in Photoshop 0 -
[quote="NNN635334001806491907" wrote:
Although I haven't asked, I'll bet they've run across the same issue and subsequently do all their sharpening in Photoshop
Why did you not ask?
I would bet that would give you a stronger knowledge of their working environment and a better chance of a good relationship and a better understanding of what a studio may find important. Useful information I would have thought.
Still, my bet would only be a guess.
On the other hand 'noise' is nearly all about the electronic engineering in the camera and sharpening suffers from attempts to curb noise. Get that fixed and I bet none of would have any problems with noise and sharpening at all.
Maybe.
Just my thoughts of course.
Grant0 -
Actually, in the cropped picture, I see most jpg artifacts in the picture that was developed with Lightroom.
Especially near edges. I couldn't live with that output....0 -
"I see most jpg artifacts in the picture that was developed with Lightroom."
Artifacts, yes, JPEG, no. Those are merely JPEG files of processed RAW data. Those are apparently LR data process artifacts. They appear to be density dependent, they're not in all transition points. This is where DxO's bokeh transition process shines (inasmuch as I didn't put it to use in this series of tests). I'm not familiar enough with LR at this time to know whether it could be addressed or not.
"Why did you not ask?"
Crazy studio schedules? I don't know where you work, but I'm surprised they ever wipe down products being shot where I freelance, never mind apply any RAW image data processes. 😉0
Post is closed for comments.
Comments
9 comments