First off I am a fan of the D800 non AA model of camera and may very well pick one up for the studio.I do not have much scientific knowledge of the AA sensor and how it removes the moire.
My question: If you were shooting a subject with a D800 with no AA, with lets say a dress that its pattern would produce moire, would it be possible to screw on a filter that would remove moire just like the AA in camera. To take it a step further, it could be a graduated filter like a CP filter
so that the subjects face is not affected by the AA filter. Then why would anyone want a D800 AA model when you could simply reach into your bag and get rid of it in a jiffy??.....and if this is a stupid question..Im sorry..the thought just popped in my head and I dont have the answer.
Not a stupid question at all, and yes you can. But you can't make it graduated.
The main problem is that it only works with fairly long lenses, 50mm would be the absolute minimum focal length where the angles of incoming light would be ok.
You also have the problem with reflections, the material compositions you use in a filter like this does not take very well to AR coatings. When the filter is situated on the sensor, it's surrounded by other layers of similar refractive indexes in direct mechanical contact, and this works like a laminated lens element inside a lens - unless the lamination breaks down it's very hard to tell that there's two different glass types in the element.
A "weather proofed" and AR coated AA filter pack would (1) be extremely expensive - up to a hundred dollar material cost, and the camera filter business LIKES having a few thousand percents of profit margin, and (2) quite thick. This would affect the lens in other ways too.
If you get a target you just simply KNOW is going to cause trouble, stop down to at least F8, F11 is better. It can be hard on the strobes depending on your setup, but generally not a bother. Just accept that medium DoF shooting does NOT blend well with moire-sensitive materials.
I've ruined product shoots by shooting at F5.6 with phase one DMF backs. It's an absolute mess when it hits bad, so bad you can never even have a distant hope of "fixing it in post". It IS unrecoverable.
But the 4.7µm pixels of the supposed D800 should respond quite a lot better to medium apertures than what the P45 6.8µm pixels (same as the Leica M9 btw) did.
I've seen a lot of threads about this (here and elsewhere) and the one thing nobody seems to talk about is the fact that Nikon will probably have recommendations for which camera to get for what circumstances. It's their sensor, they've done extensive testing, I'm sure they have the best recommendations. Personally, I'm just going to wait and see what they recommend if they really do introduce two versions of the camera.
For myself I would rather not risk having the effects of moire ruin a shoot. Stopping down is certainly an alternative but for me I love my primes and would prefer to shoot at the larger apertures.
I have heard discussion that there may be in camera processing to correct/fix moire. Now if this indeed the case I would presume that this is only going to apply to in camera processed files and not the raw files. But one certainly doesn't know yet... even if it only applies to processed files it would seem that it could be applied to the raw files if one uses capure nx2 for post work.
BTW, i have not used it extensively yet but Lightroom 4 has a moire correction brush. I used it on some light color moire I had on a D300 shot and it did an impressive job removing it.
All the moire removal tools have in common that what they leave behind is a version of the image that is so far removed from reality that a version taken with an AA filter would be on a scale factor ten closer to the "real" image.
Moire cannot be removed in post processing. Physical fact. A fact that cannot be changed. It's a signal theory chimera - given less information than what you need to solve a geometrical problem, you have to guess.
What you can do is blur the information you have, and this is what all de-moire filters do - they use algorithms meant to offset the distortion of lightness detail you get by blurring colour. In a situation where the camera would need an AA filter to get it right from the start, you are well and truly f*cked - the information lost can never be "recalculated" so that it mirrors the reality that was in front of the lens..
I made a small animation to illustrate moire in some classes, and it's quite easy to see what happens. On the top left you have the Bayer filter pattern for the 'red' channel, on the left you have a pattern that will give moire if you shoot it with a good lens, without AA filter.
Below you have what happens in reality when this image pattern is shifted across the sensor. This is what the sensor "sees", and also what gets recorded in the raw file.
The resulting [Bayer] x [pattern] = [new pattern] has got infinite variations in in shape, angle and intensity. And there's no way to know "after the fact" what really went on in front of the lens just by looking at the [new pattern] in the recorded data.
Sorry, forgot to make scales equal on that one, the large patterns on top are 200% scale versions of the patterns overlayed in the bottom, animation part.
The small strip of lines on the top end of the animation, under the "Move image pattern..." text is the scale of the image pattern overlayed on the pixels. The resulting pattern is 6x larger than the image pattern, and at another angle.
Note that the pixels are standing still - only the image pattern moves.
I'm totally confused and get dumb every time i've read anything about AA and non AA, can anyone simply tell me which is best for outdoor shooting and how much does it effect sharpness?
I have 2 questions.
1) what is the physical form of an AA filter? Is it essentially a piece of frosted glass that sits in front of the sensor? Finely sandblasted, or what?
2) with no AA filter, would deconvolution sharpening work better in theory to recover diffraction blur? Since you now only need to correct for the lens PDF and not the combined effect of lens and AA filter? You don't actually "lose" sharpness due to diffraction, rather the information is just delocalized over multiple pixels. Would this help any in recovering moire? I forsee diffraction softness potentially being a significant issue since the pixel pitch is small.
Uzay- the problem is that there is no simple answer to such a question. It really depends on what you shoot. Moire problems come up when you shoot regular, repeating patterns that have a dimension close to that of your sensor grid (or, arguably, also a dimension close to an integer multiple of your sensor grid). So windows on distant buildings, cloth, patterns in surfaces, etc. If your subjects don't contain lots of regular, periodic patterns, then you shouldn't have much issue with moire.
theSuede wrote:
What you can do is blur the information you have, and this is what all de-moire filters do - they use algorithms meant to offset the distortion of lightness detail you get by blurring colour.
Did you watch the vimeo video I posted in post #9? He didn't blur anything. He did hue and saturation changes to both the color channels and the luminosity channel. From what I could see, he didn't lose any detail in the original image like other Gaussian blur methods that are out there. Now, he does mention in the tutorial that this was just the "basic" method of removing moire and that there is an "advanced" method he uses, but he doesn't have a video of that up just yet.
Regarding the D800E (AA filter removed), it looks like it would be nice for someone who spends a lot of time post processing their images and needs to get the absolute best image quality money can buy. However, for photographers who have to work with deadlines and can't afford to spend hours editing a single image, sticking with the regular D800 with the AA filter is probably your best bet. I mean really, how many people are complaining that they just can't get sharp enough images with their current camera gear? I have seen some amazing detailed and sharp images here from people with "low-rez" 12mp cameras that still have their AA filters in them. Now all of a sudden a camera without an AA filter is coming out and people all of a sudden think their current cameras are junk. Funny how they were just fine 6 months ago.
Moire shows up as spikes in the (2D) frequency spectrum, at fixed locations. It can even be auto-detected with simple correlation. Attenuate the spikes and the moire is gone. Visual information is predominantly in the phase component (the domain of the edge function) so attenuating a few very narrow frequency bands according to a device-specific progression series has little to no other visual impact. It simply removes anything that looks like moire, whether it's actually moire or not. (And how often do you actually see anything like it in reality?) This is the complete opposite of auditory information where frequency is more critical than phase.
My thinking is that if you can control what you shoot, such as in a studio, then no AA would probably work. If you are a general/outdoor shooter like me, you will be better off with the AA filter. If not there, the lack will bite you hard from time to time.
If you get a target you just simply KNOW is going to cause trouble, stop down to at least F8, F11 is better. It can be hard on the strobes depending on your setup, but generally not a bother. Just accept that medium DoF shooting does NOT blend well with moire-sensitive materials.
I've ruined product shoots by shooting at F5.6 with phase one DMF backs. It's an absolute mess when it hits bad, so bad you can never even have a distant hope of "fixing it in post". It IS unrecoverable.
But the 4.7µm pixels of the supposed D800 should respond quite a lot better to medium apertures than what the P45 6.8µm pixels (same as the Leica M9 btw) did....Show more →
What is also important is that the updated Nikon Capture NX2 software will have a versatile Moire tool. If you do have a moire problem with the D800E, and if you shoot RAW, you will be able to fix it. My additional comments about both cameras (and the question of, Are the Pixels Now TOO Small) is at http://www.marketnews.ca/LatestNewsHeadlines/Preview:NikonD800andD800E36.3-MegapixelDSLRs.html
Tommy_D wrote:
Did you watch the vimeo video I posted in post #9? He didn't blur anything. He did hue and saturation changes to both the color channels and the luminosity channel. From what I could see, he didn't lose any detail in the original image like other Gaussian blur methods that are out there. Now, he does mention in the tutorial that this was just the "basic" method of removing moire and that there is an "advanced" method he uses, but he doesn't have a video of that up just yet.
Regarding the D800E (AA filter removed), it looks like it would be nice for someone who spends a lot of time post processing their images and needs to get the absolute best image quality money can buy. However, for photographers who have to work with deadlines and can't afford to spend hours editing a single image, sticking with the regular D800 with the AA filter is probably your best bet. I mean really, how many people are complaining that they just can't get sharp enough images with their current camera gear? I have seen some amazing detailed and sharp images here from people with "low-rez" 12mp cameras that still have their AA filters in them. Now all of a sudden a camera without an AA filter is coming out and people all of a sudden think their current cameras are junk. Funny how they were just fine 6 months ago....Show more →
For your first part of the post:
I've known about that method for a long time, and I have actually written the code base in some (very sucessfull) software packages that uses this type of RGB after-the-fact type of processing. The downside is: YOU'RE STILL LOOSING ALL TOUCH WITH REALITY! None of the finer patterns within the coarse patterns will have the correct contrasts, and most of the details within the finer pattern will be displaced, i.e not where or as they were in reality.
You have:
1) Removed all real color information, and "repainted it" with something similar - step one
2) Removed all reality based contrasts by adjusting RGB ratios until large scale luma patches are flat - step two
How does this "fix" the problem? You've basically painted in a new color, and done g*d knows what to the detail contrast. In the video, he turned a nice double-weave blue and white alternating thread cloth shirt into what basically looked like a knitwear material. How can that be "fixing" something?
