Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Resizing up in ACR/LR/Photoshop  (Read 8327 times)
digitaldog
Sr. Member
****
Offline Offline

Posts: 9225



WWW
« on: June 28, 2011, 01:11:12 PM »
ReplyReply

Just wondering if I need a reality check or my finding for a recent seminar are sound.

I took two raw image originals and wanted to see the benefits of upsizing in ACR/LR/CS5. Both images are from a Canon 5DMII.

In ACR, I can only go from the native resolution (21MP or 5616x3744) up to 25.2MP or 6144x4096. Not much if an upsize that that’s all I can do here. I compared that one upsized version to a resampled image using Photoshop’s BiCubic Sharper. I can’t see the difference on screen but using the Apply Image command to subtract the two, there is a difference.


But where things get real interesting is using Lightroom which doesn’t limit the resizing. I took the same image and sized it 250% to 14,040x9360 and did the same in Photoshop for a native 5616x3744. The differences were huge! I made sure I had output sharpening off in the Export module. I tried this on two images (one is a flat, resolution target shot on a tripod with a 100 macro).



Note that I tried using BiCubic Sharper which some have said is ‘better’ than smoother which I’ve found difficult to believe and in these two examples confirmed. BiCubic Smoother was superior but can’t hold a candle to Lightroom. Reality check, others seeing similar results?

Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3874


« Reply #1 on: June 28, 2011, 01:43:31 PM »
ReplyReply

Note that I tried using BiCubic Sharper which some have said is ‘better’ than smoother which I’ve found difficult to believe and in these two examples confirmed. BiCubic Smoother was superior but can’t hold a candle to Lightroom. Reality check, others seeing similar results?

Hi Andrew,


LR undoubtedly uses a sharpening step after the resampling, and it's resampling is called 'adaptive' (maintaining edge detail while keeping smooth areas from being sharpened). So for a fair comparison it would be better to compare the Bicubic Smoother result only after sharpening it as well.

Cheers,
Bart
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9225



WWW
« Reply #2 on: June 28, 2011, 01:55:48 PM »
ReplyReply

LR undoubtedly uses a sharpening step after the resampling, and it's resampling is called 'adaptive' (maintaining edge detail while keeping smooth areas from being sharpened). So for a fair comparison it would be better to compare the Bicubic Smoother result only after sharpening it as well.

Yes, it appears to be sharpening as it upsizes (even though I was sure to turn off output sharpening). And yes, I’d expect Bicubic Smoother to be the algorithm to use (which is tested below). I didn’t expect to see such a dramatic difference however. LR appears much better although I could now apply sharpening after the upsize in Photoshop.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
digitaldog
Sr. Member
****
Offline Offline

Posts: 9225



WWW
« Reply #3 on: June 28, 2011, 02:49:43 PM »
ReplyReply

I downloaded the demo of Perfect Resize and ran the same tests using its default settings. Then printed out an 8x10 section of what would be a 32x46 inch at 300PPI. The LR print is vastly superior. No I probably should play with the USM in Perfect Resize as well as Photoshop. I did one upsize with the USM and Noise on default and it looked awful I think due to the noise. I’ll try USM and another print. But LR is doing a fantastic job...
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3874


« Reply #4 on: June 28, 2011, 03:56:55 PM »
ReplyReply

I’ll try USM and another print. But LR is doing a fantastic job...

I'm not a fan of Perfect Resize, I prefer PhotoZoom Pro for dedicated upsampling where edge detail is important. In both cases however, it is very easy to go overboard with the settings, which results in a disconnect between edge detail and structural/micro detail.

A combination of Bicubic Smoother, and Smart Sharpening (lens blur advanced) instead of USM, or deconvolution sharpening goes a long way, and can/should be applied after resampling.

Cheers,
Bart
Logged
John MacLean
Full Member
***
Offline Offline

Posts: 169



WWW
« Reply #5 on: June 28, 2011, 08:05:09 PM »
ReplyReply

Just curious, Is the capture sharpening in LR or ACR set to Amount: 0 on all of these tests? Thanks, JM
Logged

KeithR
Sr. Member
****
Offline Offline

Posts: 632


« Reply #6 on: June 28, 2011, 08:39:50 PM »
ReplyReply

Interesting! Makes me want to tune into your Resolution Webinar this weekend.
Logged

The destination is our goal but it’s the journey that educates us.
Farmer
Sr. Member
****
Offline Offline

Posts: 1631


WWW
« Reply #7 on: June 28, 2011, 09:53:57 PM »
ReplyReply

Clever marketing by Andrew ;p ;p
Logged

Schewe
Sr. Member
****
Offline Offline

Posts: 5532


WWW
« Reply #8 on: June 28, 2011, 10:14:05 PM »
ReplyReply

Just wondering if I need a reality check or my finding for a recent seminar


Yes, your findings are consistent with the tests I've done...Lightroom (and ACR but which is hamstrung due to ui limitations) is a REALLY good place to resample-I agree...better than Photoshop (sadly) and really good even when compared to some 3rd path apps.

Between Eric AND Thomas, you've got some bright boys...which is why LR is such a good place to process/print from.
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9225



WWW
« Reply #9 on: June 29, 2011, 08:44:41 AM »
ReplyReply

Just curious, Is the capture sharpening in LR or ACR set to Amount: 0 on all of these tests? Thanks, JM

Nope, and that’s exactly the reality check I needed (thanks). The differences now are not anywhere as dramatic which is interesting in how LR uses a capture sharpening for resizing. At zero for both tests, the LR resample is still visually cleaner, but not anywhere as dramatically different. Now I have to print both. I think on output, the differences will be hardly visible.

That said, using capture sharpening and resizing produces a quite visible better result on print. Just to clarify, both rendered images had capture sharpening applied (the same amount). But presumably when LR sizes up, its doing something quite different than Photoshop from the ‘same’ capture sharpened specified raw. So I plan to show both examples. The bottom line for me as of this moment is that with capture sharpening applied in LR to produce a nice visible effect, when you size up 250%, you get a really good result on a print compared to rendering out that data, resampling later in Photoshop. Now the question is, how much post processing in Photoshop is needed to get the two closer? One process is far less work!
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3874


« Reply #10 on: June 29, 2011, 01:48:26 PM »
ReplyReply

Now the question is, how much post processing in Photoshop is needed to get the two closer? One process is far less work!

I use a PS action to create a sharpening layer, set it to luminosity blend-if mode, and sharpen it with a plugin (FocusMagic). So one click to run the action and then select the best deconvolution sharpening radius is all it takes me.

Cheers,
Bart
Logged
kikashi
Sr. Member
****
Offline Offline

Posts: 4088



« Reply #11 on: June 30, 2011, 03:17:27 AM »
ReplyReply

I use a PS action to create a sharpening layer, set it to luminosity blend-if mode, and sharpen it with a plugin (FocusMagic). So one click to run the action and then select the best deconvolution sharpening radius is all it takes me.

Cheers,
Bart
I'll bet that second phase of the process is non-trivial, though, however good your eye.

Jeremy
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3874


« Reply #12 on: June 30, 2011, 03:35:00 AM »
ReplyReply

I'll bet that second phase of the process is non-trivial, though, however good your eye.

Hi Jeremy,

Actually FocusMagic makes it very easy. A small trick I use is to bump the amount to 300%, and then increase the radius starting at zero while looking at the small preview. When you go too far up, the sharpening suddenly looks very weird (double contour width) and then you step back one notch and reset the amount to 100 or so. It's much easier to do than to describe.

It's a real shame that development of FocusMagic seems to have halted, so let's hope that the TopazLabs people get around to improving their InFocus plugin soon, because it shows great promise but needs some refinement.

Cheers,
Bart
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9225



WWW
« Reply #13 on: June 30, 2011, 02:58:16 PM »
ReplyReply

Well what I’m learning here is that whatever you use to sample up, proper capture sharpening is key to final quality when you output sharpen and print (kind of a bug duh, sorry). But if you just set a default capture sharpen in LR (or none), export at native rez and use any number of processes (I tested Photoshop, Perfect Layers, step interpolation etc), nothing beats LR’s export up with appropriate capture sharpening. All the various techniques produced roughly the same results sampling up 250%. On a print, the differences were virtually invisible for all the trouble. With LR, set a good capture sharpening, export and size up, you’re pretty much done. I did apply output sharpening for the print of course.

Oh, I did feel that when applying capture sharpening in LR, zooming in 4:1 gave me a better indication of what I’d get out the back end in the interpolated rendering rather than working at 1:1. Another big duh I suppose.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
bjanes
Sr. Member
****
Online Online

Posts: 2864



« Reply #14 on: July 01, 2011, 10:21:11 AM »
ReplyReply

Hi Jeremy,

Actually FocusMagic makes it very easy. A small trick I use is to bump the amount to 300%, and then increase the radius starting at zero while looking at the small preview. When you go too far up, the sharpening suddenly looks very weird (double contour width) and then you step back one notch and reset the amount to 100 or so. It's much easier to do than to describe.

It's a real shame that development of FocusMagic seems to have halted, so let's hope that the TopazLabs people get around to improving their InFocus plugin soon, because it shows great promise but needs some refinement.

Cheers,
Bart

Just for fun, I took the DigitalDog's image and applied sharpening with Focus Magic as Bart suggested. IMHO, I prefer the result with Focus Magic to the LR, which appears somewhat over sharpened with halos present around the sticks on the lower right portion of the image. The FM results also appears smoother and has less noise in the window areas. Sorry for the type, the caption should read Focus Magic.

Regards,

Bill


Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9225



WWW
« Reply #15 on: July 01, 2011, 10:26:00 AM »
ReplyReply

But did you print it? That’s key. LR’s capture sharpen and upsize doesn’t look that great on screen. But sizing up a5DMII capture 250% is for one reason; a big print. All my testing was output 8x11 (a section) of a much larger print.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3874


« Reply #16 on: July 01, 2011, 12:23:33 PM »
ReplyReply

But did you print it? That’s key. LR’s capture sharpen and upsize doesn’t look that great on screen. But sizing up a5DMII capture 250% is for one reason; a big print. All my testing was output 8x11 (a section) of a much larger print.

Hi Andrew,

Do you expect 2 similar (at on screen magnification) looking images to print differently? I don't.

So, the conclusion that presents itself is that both Photoshop and Lightroom can produce similar looking output. There are some differences (e.g. noise amplification, and halos) that may or may not be significant at larger output sizes.

Cheers,
Bart
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9225



WWW
« Reply #17 on: July 01, 2011, 06:07:18 PM »
ReplyReply

Do you expect 2 similar (at on screen magnification) looking images to print differently? I don't.

I don’t either but I also don’t know if they represent anything useful in terms of the process and the final print. The proof is in the print, not what you see on the screen.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
N Walker
Sr. Member
****
Offline Offline

Posts: 301


WWW
« Reply #18 on: July 12, 2011, 05:54:42 PM »
ReplyReply

Just wondering if I need a reality check or my finding for a recent seminar are sound.
 Reality check, others seeing similar results?

Lightroom uprez results in fractionally more edge acutance and slightly less noise (in dark areas) than CS5s Bicubic smoother. In large print, the very slight advantage to Lightroom is barely detectable to the most discerning eye, even with knowledge of how the comparison prints were made. It is a close shave between Lightroom and CS5 Bicubic - in fact the edge acutance from Bicubic is fractionally stronger - on screen, not detectable in print.

Attached image results are typical for me, whether capture sharpening is applied or not. On this stock image capture sharpening was applied 35 - 0.7 40. Luminance 11 - 80 - 50. I include a 100% crop guide at native resolution showing typical image detail before resizing.
« Last Edit: July 13, 2011, 02:51:50 AM by Nick Walker » Logged

Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad