[normhead]

Yeah, but he was responding to your claim that side-by-side the image that comes from a lens the "did it right" rather than post-corrected will be better. So how did come to that conclusion? What side-by-side image files are you looking at? (Point being is although you have a different conclusion, you'd need the exact same set of evidence to show support for your assertion.)

Ya, we're probably both just blowing smoke. So I guess the prudent thing to do is to withdraw both statements until someone knows what they're talking about. While I have an uncorrected DA 10-17 that I could probably get to give me some decent CA and a heavily corrected Sigma 8-16, I don't have Lightroom, so I can't help out.

Lateral CAs (color shadows at harsh contrast transitions) can be quite a problem with the DA 10-17mm Fisheye. At 10mm the lens produces a quite excessive amount of border CAs (2.5+ pixels) whereas the situation isn't quite as bad at 17mm (~1.3 pixels). Please note that lateral CAs are correctable via imaging tools.

It's interesting to have a look at the lateral CAs (color shadows at harsh contrast transitions) because this is where the new FLD elements are supposed to improve the situation. Previous Sigma ultra-wide lenses had quite some problems here but the AF 8-16mm f/4.5-5.6 DC HSM is indeed a good performer. Lateral CAs are very well controlled across the range. At an average width of around 1px at the image borders the issue may be noticeable at 100% image magnification but it's not really an issue anymore on prints.

http://en.wikipedia.org/wiki/Chromatic_aberration

In reality, even a theoretically perfect post-processing based chromatic aberration reduction-removal-correction systems do not increase image detail as a lens that is optically well corrected for chromatic aberration would for the following reasons:
Rescaling is only applicable to lateral chromatic aberration but there is also longitudinal chromatic aberration
Rescaling individual color channels result in a loss of resolution from the original image
Most camera sensors only capture a few and discrete (e.g., RGB) color channels but chromatic aberration is not discrete and occurs across the light spectrum
The dyes used in the digital camera sensors for capturing color are not very efficient so cross-channel color contamination is unavoidable and causes, for example, the chromatic aberration in the red channel to also be blended into the green channel along with any green chromatic aberration.

There you go. I'd still like to see real world examples, but that will have to do for now. We do have a few lens engineers on the site, maybe one of them will join in.

---

Last edited by normhead; 12-20-2013 at 02:59 PM.

[jimr-pdx]

To me sharpness is like oats in an oatmeal cookie. Sure it's what makes it an oatmeal cookie, but without the other ingredients it will be something other than a cookie!

[civiletti]

I just found the thread, so I may have missed a mention of this.

Besides bokeh, if you are into that, the one lens characteristic difficult to improve in post production is resolution, not sharpness. Sharpness is an edge effect that can be increased, albeit with risk of creating artifacts. Micro-contrast, also often taken for sharpness, can also be increased.

[vonBaloney]

... is resolution, not sharpness. Sharpness is an edge effect that can be increased, albeit with risk of creating artifacts. Micro-contrast, also often taken for sharpness, can also be increased.

True, but most people aren't that precise when talking about it and mean all of those things.

[monochrome]

Man, this is starting to sound like a bunch of Pension Consultants arguing about correlation and co-variance and the effect of survivor bias when adding non-correlating asset classes on the shape and elevation of the efficient frontier in portfolio construction.

And whether Zephyr or Factset has a better Style-based simulation engine in 100,000 repetition 40 year random sample back-testing routines.

[normhead]

Man, this is starting to sound like a bunch of Pension Consultants arguing about correlation and co-variance and the effect of survivor bias when adding non-correlating asset classes on the shape and elevation of the efficient frontier in portfolio construction.

And whether Zephyr or Factset has a better Style-based simulation engine in 100,000 repetition 40 year random sample back-testing routines.

I have no idea what my position is on that, give me a hint….

[vonBaloney]

Let's do a Monte Carlo simulation! Markov chains will show the way...

[northcoastgreg]

So I assume , that you have images taken with a corrected lens, right off the camera to share, and an another image, taken with a similar but less corrected lens, that needed correction, that has been corrected in light room to compare it with?

No, that's not what I mean. The only way to test the effectiveness of PP on CAs is to use the same image from the same lens. Once you introduce different lenses you've lost all control of your variables. Generally speaking (although it's hardly true universally), lenses that are better corrected for CAs will also be better all in many, if not most, other respects. When lenses are made toward a specific price point, there are tradeoffs involved, and oftentimes if the lens designer must choose between either more resolution and microcontrast on the one side, and less corrections of CA on the other, they will choose the microcontrast and the resolution. This is basically the philosophy behind Panasonic lens designs. Panasonic doesn't bother much with either CAs or even distortion, preferring to fix all that in the camera. Even the RAW files are corrected. Software-based CA correction can be quite effective.

Consider the following image taken from the DA 10-17:



Now let's look at a 1:1 view of the image uncorrected for CA:



You can see come green and purple CA along the edges of the snow bank, and some very slight CAs on some of the edges of some of the rocks. Here's the version corrected in Lightroom:



I can't see any difference between these two versions, other than the CAs are gone in the second crop. Perhaps if you pointed an electron microscope at these images, there might be a palpable difference; but that's not how these images are meant to be viewed.

[civiletti]

True, but most people aren't that precise when talking about it and mean all of those things.

That is okay in a simple discussion when everyone knows what is meant by "sharpness" of a lens, but when someone without great familiarity reads or hears about it, then finds a sharpness adjustment in a PP program, they can become confused.