Software distortion correction works - here's the proof

Started Jul 7, 2013 | Discussions thread
EEmu
Forum MemberPosts: 69
Like?
That experiment is invalid (and this issue is silly besides)
In reply to knickerhawk, Jul 8, 2013

While your experiment is interesting, it's invalid and biased to lower quality. You took an interpolated and quantized image, and then re-interpolated and quantized it twice. (Unless I'm misunderstanding your post.)

The thing people forget is that the pixels recorded by the imaging sensor are already substantially analog... There's the Bayer filter, of course, along with diffraction effects blurring the edges of the pixels at any aperture, etc. The camera creates three mathematical surfaces and then samples, quantizes, and possibly dithers them into a proper RGB pixelated image. That process is lossy (even with a 16bit full sized image), so each time you process the image you are working on something with progressively less information than was available in the RAW.

Unfortunately, there isn't really an experiment possible that proves a difference either way. In order to come close, you'd have to design two lenses, one with and one without distortion correction and compare the results. (You can't just throw in an element to correct distortion; it would work poorly and you'd be adding back in all the single lens aberrations: chromatic, spherical, coma, field curvature, astigmatism etc.) However, it's quite apparent at that point you are comparing two entirely different lenses and the exercise is quite moot!

(Here I was going to go into distortion theory and why this is all actually very insignificant, but it was getting long and detailed and not worth it.)

Why this is all ridiculous: (not addressed to the OP)

Regardless of whether or not distortion is corrected in software or optics the only thing that matters is the final performance. A lens with awful sphericals, coma, field curvature, astigmatism, etc will look incredibly soft even with perfect 0% distortion. So for example:

Oly 17mm f/1.8: Distortion: -4.54%, Corner MTF50 at f/2.8: 2134
Oly 60mm f/2.8: Distortion: 0.60%, Corner MTF50 at f/2.8: 2114
Oly 75mm f/1.8: Distortion: 0.44%, Corner MTF50 at f/2.8: 2759
Oly 15mm f/8.0: Distortion: -1.68%, Corner MTF50 at f/2.8: 1383

A high distortion lens (17mm) has the same corner res as a low distortion lens (60mm) which has much lower corner resolution than a different low distortion lens (75mm). All of which blow the pathetic but middle distortion 15mm body cap out of the water. A cheap lens you say? Sure, but the point remains that distortion characteristics aren't driving resolution here.

Further I'll point out the following: The Pan 25mm f/1.4 has -2.73% distortion, which is on the worse side. Its corner MTF50 values look like this:
f/1.4: 1660
f/2.0: 1746
f/2.8: 2213
f/4.0: 2213
f/5.6: 2233

Distortion doesn't change with aperture, so clearly distortion is not what is limiting this lens when wide open. Additionally, note that it's sharper at f/2.8 than the low distortion Oly 60mm, but the latter peaks around f/5.6 with a corner MTF50 of 2454: better, at that point, than the Pan 25mm.  Any role distortion is playing is pretty clearly insignificant.

Claiming that a lens is bad because it relies upon software corrected distortion is ridiculous. Such correction is handled completely transparently and requires the user to go out of their way to even see it. It's similar to claiming that a good legacy lens is in fact crap because it's not telecentric and produces significant vignetting on a digital camera.  What?  It wasn't designed for that?  Irrelevant!  After all, m43 lenses are designed to be used with automatic distortion correction but apparently that doesn't matter.

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