Influence of digital corrections on fine details

[mristuccia]

Hi all,

by working with Phocus on my CFV-50c files I noticed that there is a tiny deterioration on the very fine details when "lens correction" is enabled. I can notice this only when pixel peeping. Digital corrections get rid very well of chromatic aberrations, vignetting and distortion, but it seems that those advantages are counterbalanced by a tiny lack of micro details. If I think of it, I find it logical that some pixel stretching and consequent interpolation produces a small lack of the original information.

Is this a common experience here? If yes, and provided that you're not working on critical architecture stuff, do you prefer to avoid digital lens corrections and achieve the maximum fine details? Or rather the opposite?

 

[DB5]

Hi all,

by working with Phocus on my CFV-50c files I noticed that there is a tiny deterioration on the very fine details when "lens correction" is enabled. I can notice this only when pixel peeping. Digital corrections get rid very well of chromatic aberrations, vignetting and distortion, but it seems that those advantages are counterbalanced by a tiny lack of micro details. If I think of it, I find it logical that some pixel stretching and consequent interpolation produces a small lack of the original information.

Is this a common experience here? If yes, and provided that you're not working on critical architecture stuff, do you prefer to avoid digital lens corrections and achieve the maximum fine details? Or rather the opposite?

yes, that is the same with all digital correction. I leave mine off where possible for that reason and correct where absolutely necessary. To be honest I find degradation in file quality even from colour, contrast, exposure adjustments and such. Even something as simple as rotating the canvas or layer in Photoshop negatively affects image quality. The absolute least you can do the better.

 

[sog1927]

Putting my software engineer hat on for a minute: Is this still the case if you just turn off distortion correction? It seems to me that distortion correction would have the greatest potential for this because you're actually moving pixels around to do the correction. I'm just wondering if you might be able to retain the other corrections while preserving the fine detail. Since you can independently enable/disable the distortion correction, this would seem like an interesting test.

 

[dougpeterson]

Digital correction (e.g. lens distortion) in any software always reduces image quality by a non-zero amount. But please do not assume that all software packages behave equally in how much the quality is reduced.

Depending on the amount of correction (correcting the distortion from a lens with 0.5% distortion is very different than correcting distortion from a lens with 5% distortion) and the software used (the math in different raw processors is not equal, and some raw processors are very clear leaders here) the amount by which the quality is lowered can be pretty meaningful or can be absolutely inconsequential.

So when testing such things, make sure both the software and the lens are considered as variables.

Given that this thread is about a brand that my company does not sell I will leave it at that.

 

[mristuccia]

Thank you all for your feedbacks!

As a software engineer myself, I do think too that vignetting correction, for example, should not create a loss of fine details, and that distortion correction is the best candidate for the worst impact on that side. I will try corrections separately in order to prove that.
Beside this, at this point I too will the avoid digital corrections unless it is really necessary.

@Doug:
I agree with you that an important factor is the original distortion of the lens itself. I tend to agree less on the fact that there can be softwares which do a better work than others. At least if we deal with the leading RAW developers out there (C1, Phocus, LR), and a distortion correction can be made only in one way. That is by stretching pixels. There are well known and well described algorithms in the IT literature for this. Not that much space for personal brand's "magic" over there. IMHO.

Kind regards,

 

[dougpeterson]

There are well known and well described algorithms in the IT literature for this. Not that much space for personal brand's "magic" over there. IMHO.

You might want to check that assumption with actual testing; you'll be surprised . The math for lens distortion is only one part of a total chain of math applied between [raw data] and [final pixel] and its how the chain reacts in its totality that matters to the final image.

 

[mristuccia]

You might want to check that assumption with actual testing; you'll be surprised . The math for lens distortion is only one part of a total chain of math applied between [raw data] and [final pixel] and its how the chain reacts in its totality that matters to the final image.

True. But I doubt that, given a lens distortion algorithm enabled, another node of the chain will counteract the loss of fine details that the former (or the latter, depending on the chain order) introduces. It can only get worse, maybe to a greater or lesser extent depending on the quality of the chain and its reaction, as you stated with my full agreement.
But I'm talking about the original sin here. The one that can't be remedied anymore.

Best,
Marco