AJSJones
Offline
• • • •
Upload & Sell: Off
|
 p.1 #11 · R5 vs 5DSR Landscape and the AA filter - DPR has comparisons | |
The Nikon and Canon "AA-less" filters (D800e and 5DsR) are both "canceled" style. The d800 and D800e had the same filter stacks except in one the second AA filter was rotated around the optical axis relative to the "normal" orientation and undid the effect of the first (brought the two dots back together instead of making 4 dots). Filter stack thickness is a critical parameter (LensRentals blog has some info on that here), so having a thinner stack may be detrimental overall. I think if the R5 had the new 16 point AA filter we would have heard about it.
In any case, a significant amount of the effect of the AA filter can be reversed in PP - its PSF is known and decolvolution sharpening is quite fast these days. Even the old 200,0.3,0 sharpen routine recommended by Chuck Westfall returns a lot. I was curious a long time ago, so these examples don't use deconvolution but showed, to me, at least, the effect of the 1 pixel AA was not as bad as I had expected and that much could be recovered. These are 100% crops: start with RGB image, duplicate layer, displace one pixel up, opacity-0>50%, flatten, duplicate, displace one pixel left, set opacity 50% flatten, to simulate the AA, then sharpened. (Ideally this would have been done on the "un-de-mosaiced" file but the effect is clearly much less severe than the 1 pixel Gaussian blur.)
This article specifically addresses increasing the sensor resolution and how different lebneses perform and includes a useful evaluation of the effects of AA filters (yay, with pictures!) The loss due to AA filters is detectable but not severe, and much can be recovered. Using numbers like 5% or 10% "resolution loss" is a little too simplified/ill-specified to do the right comparisons. There is a comparison towards the end of the article of an "excelllent lens" on sensors with increasing resolution with and without an OLPF and it's hard to evemn see the curves shift at the higher spatial frequencies. It's a great article and worth the time. DigiLloyd showed some deconvolution sharpening examples
here a while ago. (I might be able to get some Topaz InFocus - a form of deconvolution) examples when I get back to my desktop
Bottom line - OLPFs are not as bad as many people "think". Of course, I'll let you decide, once you've compared deconvolution sharpening on an AA image with a non-AA image (which can't take as much PP sharpening) whether that small residual difference is a dealbreaker, even at your print size It may be, but it's only one factor in your decision making process
(Full disclosure, I have a 5DsR and those same lenses you use for "roadside" venues, although my printer is only 24" so my prints are 24x 30, 36, 48 and 60, and if the 5R came with a canceled AA filter as an option at a premium, I would probably decline these days, since I have "mastered" the Topaz InFocus sharpening plug-in and my prints are still generally printer resolution-limited rather than image pixel quality-limited -often stitched).
|
