Resolution of M43 lenses

Started Mar 31, 2013 | Discussions thread
Anders W
Forum ProPosts: 17,055Gear list
Like?
Re: Demonstration of Differences in DxO Geometric Corrections
In reply to Detail Man, Apr 3, 2013

Detail Man wrote:

Anders W wrote:

Detail Man wrote:

Anders W wrote:

Sure. DxO crops the image to a rectangular shape. If you ask LR kindly, I think it will do the same. Or if you ask DxO kindly, it might abstain from the cropping, which from a diagnostic point of view might be even better. If additionally, you make sure that the amount of correction applied by the two converters is exactly the same, it might be an interesting comparison.

OK, magister, I have now cropped the Adoobie image down to the girlie-man size of a mere 3499x2624 pixels (seemingly all that it could handle), and gratiously lowered the DxO Optiques Prose image size to a similar puny dimension. Once again, download the Original images and display them in an image-viewer at 20% size (collapsing 5x5 pixel arrays into single, centered pixels).

Lightroom 3.6:

Note the pattern that appears to imply a single centrally placed optical correction reference point.

.

DxO Optics Pro 7.23:

Note the four circular patterns that appear to imply four separate correction reference points.

I anxiously await your refutations showing that Adobe is the embodiment of crystalline perfection.

Note: The DPReview post-display system appears to be malfunctioning. This is my 2nd and final possible edit. If images do not display, click on the Gallery Page links to download the Originals.

DM ...

I see all kinds of interesting moiré patterns depending on the magnification of the image itself as well as display magnification. But I wouldn't want to make anything out of that.

I grant you that reasonable uncertainty (of the appearance of patterns, even in the case of decimation ratios of 5, where we know that 5x5 pixel-arrays collapse into a single display pixel).

Now if you really want to test this DM, one way of actually getting somewhere without a whole lot of work would be the following: Take your test image, distort it using the one RAW converter, output it, and then try to undistort it using the other RAW converter. If the algorithm is essentially the same, you should be able to come back to square one, i.e., the pure, straight, column-row pattern that you started out with (though perhaps a bit degenerated in ways that are for present purposes a bit irrelevant). If not, then you shouldn't. For safety's sake, you might want to approach the matter from both ends, distorting with LR and undistorting with DxO, distorting with DxO and undistorting with LR.

That procedure would work in the case that both processors are known and understood to accomplish all geometric corrections only after de-mosaicing (such that TIF manipulation is representative). Maybe, maybe not. It (may) depend on an individual processors' architecture.

I was merely trying to be helpful in finding a more straightforward and conclusive test of the matter. If you find this one insufficient, feel free to suggest an even better one.

My guess is that when you just use the main distortion slider of the respective RAW converter, they will for all practical purposes do the same job. When working on the basis of lens profiles, however, things may well stand differently. In that case, both may make use of more than a single parameter to correct more complex forms of distortion, and the parameters they use may be different.

Have posted a number of Originals that demonstrate the geometric differences that can (as far as I am concerned) be easily seen by (at least my own) eyes (of DxO geometric corrections relative to large numbers of OOC JPGs, as well as some RAW Therapee, Lightroom, and Silkypix corrections).

You seem to perceive such differences as minor, relegating them to the realm of the virtually unmeasurable, the insignificant, the subjective. However, you have not spent the time that I have looking carefully, and your perceptual processes may simply not be cognizant and/or interested in the visible differences. That's OK - but I do not see that as a basis for your speculation.

I honestly have no idea what you are referring to when you talk about my "speculation" or the basis of it. Furthermore, my subjective criteria for judging the quality of an image in this or that regard do not pretend to be more than precisely that: my subjective criteria. I don't demand that you or anyone else share them.

As I pointed out in my previous message, I think DxO distortion correction based on lens profiles may well be different in at least some cases from what LR or RT can do on the basis of a single slider control. The latter uses a single parameter whereas the profiles surely uses several unless the distortion is very clean and simple, without any mustaches or things like that. LR can, if I am correctly informed, use multiple parameters too, if you set up a lens profile to do the correction rather than just use the slider. So the appropriate comparisons here would be between LR and DxO with both using the sliders and between LR and DxO with both using good quality lens profiles.

As long as you use the distortion slider only, I doubt that you'll find any difference to speak of between DxO and LR. I also doubt that you'll see much of a difference between the quality of the distortion correction between the two converters as long as it is MFT lenses we are talking about for the simple reason that they are unlikely to show complex forms of distortion to any significant degree. For other lenses, especially those with complicated forms of distortion, the quality of the lens profile may be of considerable importance. But in this case too, I don't think there is any major difference between what the converters can do if only the quality of the profile is the same.

One would have to assume that DxO Labs goes to the trouble of characterizing individual lens-systems, then abandons that information in favor of corrections that have more "artistic appeal" to their customers than technical accuracy comporting with what they have measured.

Or, one would have to assume that DxO Labs is merely "obsessing" in implementing their optical corrections to levels of accuracy that you personally may find unecessary because your own eyes do not happen to object to what (to my eyes) appears to be somewhat simpler and (personally perceived by me to also be less effective) implementations of geometric corrections.

None of the above follows from anything I have said.

The issue of primary concern in your mind appears to be whether your particular tool of choice is adequate for your own perceptual purposes, preferences, and tastes. And, it seems that it is.

When it comes to choosing a RAW converter for myself, that is indeed my primary concern and I see nothing wrong with that. This doesn't necessarily mean that it is a primary concern in other contexts.

Note: I don't know how you create the synthetic RW2 that you are playing with. But it should obviously not be such that either RAW converter tries to apply any distortion correction by default. Any distorting or undistorting should be done by you via the sliders.

It is true that the Panasonic image-file meta-data for a LGV 14-140mm at 25mm FL exists in the RW2 test file, and Lightroom (unlike DxO Optics Pro and RAW Therapee) unfortunately does not allow the user to defeat its application in silent geometric corrections. It may be that Lightroom's manual adjustments multiply together with (as opposed to simply add to in the control-chain) such pre-applied and non-defeatable "silent" geometric corrections. The result of such a multiplicative process (may) result in composite effects that cannot later be separated.

 Anders W's gear list:Anders W's gear list
Panasonic Lumix DMC-G1 Olympus OM-D E-M5 Olympus E-M1 Panasonic Lumix G Vario 14-45mm F3.5-5.6 ASPH OIS Panasonic Lumix G Vario 7-14mm F4 ASPH +21 more
Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow