Lr3 Rendering quality with 5D2 and G10??

[tommm]

Having heard all the rave reviews from those testing the beta release I was really looking forward to the improved IQ with Lr3. With old 5D files I see a distinct improvement but with 5D2 and G10 files unless I'm imagining things there seems to be less fine detail. Less noise for sure but fine detail definitely seems less and if sharpening is increased rather than being able to sharpen until things get a little crunchy and then backing off a slightly "wormy" characteristic seems to appear without ever reaching the same level of fine detail in Lr2.7. Hope it just the few images I've looked at so far because I was hoping for IQ akin to Raw developer.

Am I imagining things or has anyone else found the same thing?

Tom

[John R Smith]

Tom

With my Hasselblad 3FR files I don't really see any improvement in IQ with LR3 over LR2. If you are running 2.7 rather than 2.6 you already have the new demosaicing algorithm anyway, apparently. The noise reduction in LR3 is certainly a big improvement, but it only matters if you shoot at high ISO. I mostly shoot at ISO 100 and 200 handheld, or ISO 50 off a tripod, and I find the NR in 2.7 is perfectly fine at the default level for my shadow areas and skies. When I convert 3FR files from PV 2003 to 2010 in LR3 and compare them at 100%, I simply can't see any worthwhile difference, no matter how much I peer at the pixels. I have debated this before with Jeff, and he says he does see a difference with his P65 files, so it is probably just down to the particular raw files that you are dealing with. For some cameras and ISO settings there will be a major gain with LR3, for others less so or not at all, perhaps.

John

[madmanchan]

Fine detail extraction should actually be much better in LR3 for the 5D2 and G10. But the sharpening algorithms have changed, so the exact numbers that you used with LR 2.x will likely have to change for LR 3.x (once you update to Process 2010).

[tommm]

Thanks. I've tried many other raw converters and agree that DPP and more so RawDeveloper have seemed to have slightly better rendering of CR2s than Lr in the past, however Lr has a far superior work flow so I've been happy to go with Lr and just do the occasional image in RawDeveloper when needed.

Lr3 has had the underlying algorithm used for rendering revamped and was reputed to give far higher IQ than 2.7 and even RawDeveloper, unfortunately so far with 5D2 and G10 files it seems if anything worse than 2.7 to me (less fine detail at the same sharpening and a "wormy" characteristic appearing when sharpening increased). Hopefully it's just a peculiar feature of the few images I've had a close look at so far and not the case for 5D2s and G10s in general. The answer I'd like to hear is lots of 5D2 / G10 owners telling me I'm talking nonsense and that it's far better than 2.7! Would be good to get some feedback from other Canon owners.

Cheers,

Tom

[terence_patrick]

Interesting that you should ask this question...

Over the weekend, I went to Yosemite National Park to attempt to shoot the "moonbow" (rainbow at night from the full moon)

I got home, loaded the pictures into LR3 and... hmm... I discovered the "banding" problem that people with 5D2s have reported when shooting pictures that contain large dark areas.

Then I thought "What about DPP? Will that be any better?"

The picture is 28mm at f/4 (24-105L), 66 secs, ISO 400, 11:28pm on 26-6-2010.

All JPEGs come from the very same CR2 file, the only difference is the renderer: LR2.7, LR3 and DPP.

With DPP, all that I did was "Convert and Save", followed by resize.
With LR, all that I did was "Auto Tone", followed by an Export.
All on  the same computer.
The small file comes from DPP, the largest from LR3, the other from LR2.7 and I believe this ordering matches how it appears at the bottom of this posting.

I wondered if perhaps LR3 was over exposing but dropping the exposure does not solve the problem; most notably, the sky colour is never the same as in DPP which implies more complex changes to the tone curve/saturation.

The most obvious issue to me is that DPP appears to know how to deal with pattern noise in dark areas in a much more comprehensive mode than does LR (be it 2.7 or 3.)

Now maybe I should sit around and spend minutes or hours tuning LR so that the image is "right" but if DPP can do that then what's the point?

How is all of this related to your question...

You're asking whether or not LR2.7 or LR3 is better for the rendering of fine detail. My answer is that perhaps you need to expand the set of software from which you are making comparisons.

What if you tried, upon import into LR3, adjusting the default brightness down to 0 and switching the camera profile to Camera Faithful or Camera Neutral? The Auto Tone thing might also be making it hard to match DPP. But I suppose if what you're after is DPP's processing, you might be spinning the wrong wheel using LR (I should know, I spent a little too much time getting LR to respond to colors the way Capture One Pro does.).

Re: the original post
I'm seeing good differences in my 5D2 and even in my G9 files from LR3, overall, but have noticed from time to time that hair details can sometimes become very jaggity, especially if the hair stands are diagonal (I shoot people, not landscapes).

[John R Smith]

There is a lot of debate generally in this Forum on the apparent strengths and weaknesses of various raw processors with specific raw file formats. For example, Phase users swear by C1, 'Blad users by Phocus, and so on. It seems to me that a lot of this debate somewhat misses the point, because the protagonists of these raw format / software combinations are judging the issue on the basis of a 100% view on their PC monitor. However, the real arbiter of quality is the final print. And this is where we have to judge results, by the A4, A3, A2 or whatever size of print in the hand or on the wall. Lightroom's print quality is truly excellent, and in most cases it seems to me that you will have a better final print using LR as raw developer, editor and print spooler despite any small differences you might think you see on screen at 100% using some other raw processor.

So to judge the difference between LR2 and LR3 I would (and do) compare the same size print from the same image file in my hand at normal viewing distance. And for my 3FR files, I can see no difference.

Which saves me 75 quid in upgrade costs  

John

[mbridgers]

Brooks Jensen from Lenswork noticed differences in LR3 and LR2 as well, and wrote about them here:

http://technology.lenswork.com/2010/06/lig...n-updating.html
http://technology.lenswork.com/2010/06/lig...-version-2.html
http://technology.lenswork.com/2010/06/mor...conclusion.html
http://technology.lenswork.com/2010/06/a-l...of-insight.html

Maybe his insights can help.