Ad
Ad
Ad
Pages: « 1 [2]   Bottom of Page
Print
Author Topic: Capture Sharpening in Lr4  (Read 6240 times)
RFPhotography
Guest
« Reply #20 on: September 27, 2012, 06:49:06 AM »
ReplyReply

Jeff, does the size of the installed user base really matter?  Isn't it more the work required to create a 'blur profile' for every camera that would be the impediment to implementing capture sharpening based on the specific sensor or sensor/lens combination that would be the problem.  There was a large, installed user base when PV2012 was introduced and that was a pretty 'radical' change.  Doesn't seem Adobe is put off by making changes that will upset (even if for a short period) a not insignificant portion of their users base.
Logged
AFairley
Sr. Member
****
Offline Offline

Posts: 1144



« Reply #21 on: September 27, 2012, 08:26:15 AM »
ReplyReply

Hi Jeff,

I am aware of human interface design issues. But that's exactly what I was hinting at, the revolution can be under the hood, no need to upset the installed user base. The blur analysis can take place without user intervention, and the result can (automatically, with an auto button) be used to preset the current controls to something more meaningful. I find it hard to believe that that wasn't considered, if one in fact already made the mental switch to a different approach to Capture sharpening. I hope I'm wrong, but that's why I suspect that that penny has not dropped yet. Maybe this discussion can help.

Cheers,
Bart

Yes, this sounds as if there possibly could be something like the recent lens cast plugin, where a reference target is used as the basis for the adjustment applied to the image?
Logged

bjanes
Sr. Member
****
Offline Offline

Posts: 2785



« Reply #22 on: September 27, 2012, 09:23:48 AM »
ReplyReply

Exactly, although better targets are already designed for such a purpose.

I would suggest something like this:

The benefit of such a target pattern is that it's easy to produce at any size (it's vector based, therefore no resampling losses), and it allows to determine slanted edge responses for any angle, not only for horizontal and vertical resolution (although those are the main factors with rectangular sensor arrays, but e.g. lens decentering requires more flexible methods). The shooting distance is not critical, although shooting it square is. It also allows to compensate for uneven lighting, and can be combined with e.g. Opto-Electronic Conversion Function (OECF) patterns such as specified in ISO 14524, to calibrate for tonecurve and gamma effects.

Bart,

That is an interesting target--an alternative to the dead leaves model? Anyway, how would one use this target? You have already presented your web based method employing a slanted edge target to determine the optimal sharpening radius. How does this relate to the new target?

Regards,

Bill
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2785



« Reply #23 on: September 27, 2012, 09:51:43 AM »
ReplyReply

Nope...I was there in the meetings where the "defaults" were derived and determined...the motive was nothing more that coming close to previous version's sharpening effects–nothing more. The thought was people would learn how to use the sliders to get what they need. Obviously, that hasn't really happened...(although I've gotten pretty good at making the optimal adjustments for my own work).

Matbe the engineers will turn their attention towards better demosiacing & sharpening (they are toed together at the hip). Maybe not. But it behooves users to learn how to use what they have now...

Jeff,

That is good advice. The deconvolution methods offer promise, but difficulties in obtaining a good PSF (point spread function) limits their use for everyday work. Bart is at the forefront in these techniques and I have followed his posts with great interest. However, in my own work I have found that they offer only limited advantage over what LR/ACR already have. The disadvantage is that one loses the marked advantages of parametric editing and the simplified printing that LR offers.

Bart does argue persuasively for setting the radius for image restoration using deconvolution rather than image content. However, at times, one does not desire every skin pore and blemish to show up in a portrait and one can make an argument for low frequency edge sharpening as you demonstrate with the portrait of a beautiful young woman in your new Digital Negative book.

Comments by you and others are welcome on the relative merits of these two approached.

Regards,

Bill
Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5469


WWW
« Reply #24 on: September 27, 2012, 12:07:20 PM »
ReplyReply

Jeff, does the size of the installed user base really matter?

I was referring to the time of the ACR 4.1 update that the new sharpening was put in. That was the first time a Process Version was devised. There was a desire to keep the new ACR 4.1 sharpening defaults "similar" to ACR 4 and earlier. Obviously, the new PV 2012 has had a major tone mapping impact and was pretty brave to do. If you hung out at the LR4 beta forums for any time you saw a lot of blowback.
Logged
RFPhotography
Guest
« Reply #25 on: September 27, 2012, 01:00:35 PM »
ReplyReply

I was referring to the time of the ACR 4.1 update that the new sharpening was put in. That was the first time a Process Version was devised. There was a desire to keep the new ACR 4.1 sharpening defaults "similar" to ACR 4 and earlier. Obviously, the new PV 2012 has had a major tone mapping impact and was pretty brave to do. If you hung out at the LR4 beta forums for any time you saw a lot of blowback.

Right, I understand that.  My point being that (a) they did it once so why not do it again, but (b) it would be a lot of work to come up with a blur profile for every sensor out there and then come up with specific default capture sharpening settings that were sensor/camera specific.

It'd be even more work if lens blur were brought into the equation and since we dealt with lens blur for over 100 years with film that's probably not a big worry despite what Bart might think. 
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3581


« Reply #26 on: September 27, 2012, 03:52:49 PM »
ReplyReply

That is an interesting target--an alternative to the dead leaves model?

Hi Bill,

Well they allow to do similar things, e.g. determine MTF curves, but they are also different in how they achieve that. The type of target that I suggested has several very interesting characteristics (sub-pixel oversampling accuracy is just one of them) that are not as available to exploit in the dead-leaves version. The symmetry allows to additionally do distortion calculations at the points where the patches intersect. A single target shot therefore delivers multipurpose data. The target also allows to check for uniformity of lighting, and by comparing the patches at different positions it is possible to judge whether the target was shot square, all usefull for test quality control.

Quote
Anyway, how would one use this target? You have already presented your web based method employing a slanted edge target to determine the optimal sharpening radius. How does this relate to the new target?

It has been researched that the slant angle of 5-6 degrees versus the horizontal or vertical axis provides the best quality for SFR/MTF calculations. The only (minor) drawback is that it restricts the outcome to 2 orthogonal axes. The proposed new pattern has edges at all 360 degrees, not just 2 orientations. This allows to produce MTFs at any rotation angle, which makes shooting the target 'rotation invariant'.

For the simpler task of just determining the PSF, it also offers different benefits. It does not only allow to determine a simplified elliptical PSF (where only the hor/ver sigma dimensions can vary), but also an even more accurate PSF for more complex blur shapes (e.g. astigmatism with a more diagonal elliptical shape towards the corners).

The required math is a bit more complex, but that's why we use computers. The same calculation principles can be used to analyse regular images which also have edge detail at various orientations. Depending on the required accuracy or calculation speed it is possible to use different calculation methods, but they all require a more complex software application than my current webtool.

Cheers,
Bart
Logged
Pages: « 1 [2]   Top of Page
Print
Jump to:  

Ad
Ad
Ad