Re: Would you prefer D800 with or without AA filter ?
Kittyk wrote:
ausemmao: you need to put more faith into humanity. We managed to land on Moon in 60ties, by 2012 we should know how to mathematically remove moire as good or better as AA filter does. Especially when we have all sensor inputs in RAWs and have time to do even little brute-force (keep testing until we find good one).
This is where understanding the theory (have to understand it for autopilot design) beats \"feeling\".
It\'s mathematically not possible to truly remove aliasing if your signal+sampling combination would create it (in the sense that you want real information) unless you have a fully sampled signal to compare (and if you have the fully sampled signal, why aren\'t you using that instead? ). It\'s inherent to discretely sampling a continuous signal.
That said, I don\'t think a fully faithful signal (image data) is the most important thing about an image. If it were, we wouldn\'t be using cameras the way we do. If it were, we wouldn\'t ever do any noise reduction beyond dark frame subtraction, because any NR beyond that destroys detail.
The picture you linked is a great example. The thread pattern in the groom\'s jacket in reality doesn\'t look like it does in the full image nor the crop. But if a person were presented with that picture, would they ever say \"hmm, the thing that breaks this image for me is that the fabric doesn\'t look exactly right\"? (notwithstanding that a person looking at the groom with their own eyes from where you took that picture would be seeing some aliasing effects ).
There\'s also a difference between \"this is something that happens to an image due to a limitation in technology\" and \"this is something that is photgraphically relevant\".
I really wonder how the discussion might have changed if MF sensors had been supersized DSLR ones rather than downsized astro/surveiilance ones.
Feb 05, 2012 at 04:15 PM
Previous versions of ausemmao's message #10312889 « Would you prefer D800 with or without AA filter ? »
). It\'s inherent to discretely sampling a continuous signal. 
).