Something to play with

[Jamie Roberts]

Advertisement (gone after registration)

Ok, I know we're all going to hand-code the CV 15 (and other CVs) with adapters, special tape, and so on.

 

But for those of you who want to play with a quick correction in PS for the cyan vignette on the uncoded CV 15, well, here it is.

 

http://www.fouldsroberts.ca/test/workaround/cyan_curve.zip

 

Heliar_CV_15_LAB.acv is a Photoshop curve file for CV15 M8 shots, meant to be used in LAB mode.

 

Here's what you do...

1. Open your shot
   
2. Edit... Convert to Profile and select LAB color. Don't just change modes, doing the conversion makes PS do some extra, good, stuff.
   
3. Add a curves adjustment layer.
   
4. Load the Heliar_CV_15_LAB.acv curve (PS CS3 this is a preset)
   
5. Adjust opacity to taste. I like to start at 50% and go up from there.

Note the curve will remove cyan pretty aggressively, as well as the vignette itself. You can do a couple of things to mess with this (at least)...

 

First, you can play with the L curve--which is the Lightness curve. Heck, if you're brave, you can mess around with the A (green-magenta channel) or B (yellow-blue channel) curves too.

 

You can also paint in the layer mask (the vignette mask) with black or white to either remove or add back cyan selectively. Of course, the mask is based on the CV 15 vignette, but you can get as creative as you like.

 

If you're on Windows, and want this automated, I happen to have a little action for you to load into your actions palette.

 

Put the Heliar_CV_15_LAB.acv into C:\Windows\ directory so the action can find the curve, and load JHR_M8_IR_Vignette.atn

 

Have fun! This should work pretty well with the CV 12 too, though I haven't tested this.

 

Personally, I can't wait to get the lens coded somehow.

 

Caveats: I spent about 15 minutes tweaking this, and the action. Is it perfect? No way, but it's a good start. Is it clever? Not by half--it's a brute force correction with a gradient layer mask. Will this replace a coded lens? No--Leica is doing this a the RAW level; post production is always going to be a little worse.

 

There are no doubt many more elegant, and effective Photoshop ways to work this, too. My apologies to the PS crew This was more an experiment for me to see what Leica is trying in firmware. YMMV. Void where prohibited. Etc... etc...

[scott kirkpatrick]

Right now the M8 firmware puts a strong L vignetting correction plus a weak A channel correction into the ON behavior, and puts a weak or nonexistent L vignetting correction and a stronger, pretty well lens-calibrated correction (see Sean Reid's results, and one of my posts about a week ago) into the ON+UV/IR behavior. When the firmware doesn't get things quite right, you can end up with slight over or under corrections of the A channel and need either a lot more help to reduce L vignetting, or small adjustments to reduce over corrections (from using ON without the UV/IR). This means that the PP requirements really separate into L vignetting and A vignetting corrections.

 

The fact that the curve shape changes -- it gets more extreme as the focal length decreases for A channel and only matters at the widest apertures for L vignetting -- means that to solve these problems we need a programmable gradient map.

 

I've managed not to learn much Photoshop. Is all this easily doable in an action?

 

scott

[Fotomiguel]

Thank you Jaime! You're great!

I'm still waiting for the 15cv. My dealer tell me, it will need some time. I thought I was going to hand coded it, but if your curve works...perhaps I don't need to do it.

Do I need PS C3? I have ps4: What about ps5?

Thank you very much!

[Jamie Roberts]

Scott--

 

Never Post at Night. That should be my new rule.

 

Yes, you have to correct less as you stop down on any given lens (because the vignette goes away), but that's why the correction curve itself should be on a gradient mask on the adjustment layer.

 

D'oh. Evidently I didn't learn PS either. How did I forget to tell people about that step!?

 

In the Windows action I posted, it's all there...including the layer mask.

 

But if you're on a Mac, to get this to work, you need to add a layer mask to the curve, click the circular gradient fill and add the mask as appropriate.

 

Anyway, I'm usually up on Sean's reviews--I'm a regular reader of his, and his site is fabulous--but I just haven't had the cycles in the last few weeks to check it out. I will. I'll also go back and read your post.

 

Having said all that, let me respond to your note here by saying that in my admittedly quick and shallow testing, I found a single L and A/B correction just fine for the CV 15 lens I was working with.

 

To me, this is not about accuracy, actually--more about "accurate enough" in a pinch.

The only issue I've seen so far is a slight colour deficit in the B channel (you need a B correction too to get rid of cyan, since to correct cyan you need a negative red correction and red, in LAB is essentially +B +A).

 

And you're right of course, you need to adjust the L and and intensity of the AB as the lens varies (and aperture varies). The variance on the CV 15 isn't so great, though, that I couldn't take care of it mostly with a quick opacity change, as opposed to a programmable gradient.

 

There's no way I know of right now to program a gradient in PS--but the program now accepts Script, so if you had EXIF data you could probably automate it.

 

The PTLens set of plugins on Windows is the closest I've seen, but of course for non-coded lenses there's no EXIF data to work on anyway.

 

So this is another YMMV! When I get a chance, I'll play with the CV 21 and 28 and see what's what there.

 

On the other hand, as I suggested elsewhere, if you were really clever you could probably detect the increase in Cyan as you move away from the center (on average) and take a stab at automatic correction.

 

Lens data--the way Leica does it--is much more reliable, of course.

 

The nice thing about working in LAB of course is that you can change colour without affecting luminance detail at all; you can't do that in any other colour space.

 

Miguel, this was done in CS3, but should work back to, oh, PS7 let's say! I have no idea how this would work in previous versions of PS (my memory isn't that good, but curves and LAB should be there still).

 

By All Means hand code your lenses! This is more of a "what if I get caught" thought-experiment on my part

[Chuck Hatcher]

Thanks for posting that, Jamie.

 

My needs may not be as critical as some, but I have already decided to forego lens coding in favor of post processing when necessary (and for me it usually isn't). Not that I wouldn't still like for Leica to add a means to manually enter lens data into the camera. I use older Leica and newer CV lenses, and given the ever-increasing cost of new Leica glass, that isn't likely to change. Unlike the IR color shifts, compensating for cyan or normal vignetting seems like something that can be taken care of very well in processing.

[grdglass]

Thanks, Jamie.

 

Helene

[Jamie Roberts]

Advertisement (gone after registration)

Chuck, Helene--you're welcome! I think it's a neat demonstration of how you can correct stuff, anyway

[jedisdl]

Thanks Jamie - I'll try it out tonight (my wife won't be happy, though when I hog the computer til I get it right.)

 

Steve

[carstenw]

Chuck, the latest LFI has an interview with Stephen Lee in which he says something like "not everyone needs 60% contrast at 40 lppm". I think that Leica lenses are likely to make different design compromises in the future in order to bring a second line of more affordable--yet still excellent--lenses to market, like the 28 Elmarit Asph and others, perhaps.

[sean_reid]

Thanks for doing this Jamie. In the tests I'm doing right now, the in-camera (16 mm) corrections work fairly well for the 15. The 12 however, as I expected, is not well corrected by the 16 mm in-camera corrections. As such, I'd love to see some folks work on an action to correct for the cyan drift seen in files from the CV 12 + IR cut filter. I've done some initial tests today using a 46-55 step up ring and a 55 mm 486 filter. The 55 is almost large enough to not vignette on the 12. In fact, with the right custom adapter it might be just big enough. Otherwise we'd need to look to an adapter that can hold a 60 mm filter. I'm sure John could create one (maybe with an integral hood as well as the 60 mm filter threads) using the same attachment method as the stock hood.

 

In any case, those of us who want to use the CV 12 with IR-cut filters will need to find an adapter-maker and then figure out the PS corrections (unless we do them in RAW conversion in C1 4.0 <G>)

 

Cheers,

 

Sean

[Jamie Roberts]

Hey Sean,

 

I don't currently have the 12mm CV, but it's something I think about every day

 

The correction for the vignette on that one should be relatively the same as the 15. I'd do a conversion to LAB, create a curves adjustement layer to neutralize the cyan in the A & B and to lessen the actual vignette in the L channel.

 

The best way of getting close to the lens characteristic is to apply a layer mask on the curves layer and draw a mask with a radial gradient from the center on out. The trick is getting the right gradient for the mask, which varies by lens.

 

(But the gradient is a quick way to do this. In fact, I'm finding I like this method for all kinds of "feathered" adjustments I'd normally do with a multi-step selection!)

 

 

And of course we'll be doing this in C1 V4 Sigh. Can always hope!