rico wrote:
Images made through a properly implemented low-pass filter are always more accurate, meaning a truer representation of reality.
sorry to beat a dead horse, but this statement is unfortunately not true. unfiltered images can be more accurate than "properly filtered images". it depends on the spectral content and periodicity of the photographed content. also, to implement a low-pass filter so that there is no false data one must filter a fair bit below nyquist which most photographers would consider too soft. thus, even with our AA filters we often get false data.
Tariq Gibran wrote:
Thanks. So, do you think with this particular subject matter the regular AA filtered D800 would have represented the hair more accurately? I guess another way to put it is: would a properly calibrated AA filter do a better job than trying to take care of moire after it occurs?
Yes. If you WANT to shoot at F5.6 > F9.0 all the time, use strobes and sometimes have the luck to score a red-head for the shoot - the normal version would be preferable.
If you HAVE the option to stop down just a stop more (F11?) for suspect cases you're way safe, even with the E version.
Look at the fabric - now we don't know if there's been any PP fiddling going on here, but it looks fine to me. But the fabric is pale yellow, and not really high-contrast weave patterns, so I doubt you'd get problems at F9 with this. There's a strong flat correlation in green to red>blue in that fabric, which means that moire will be less strong.
As I've said repeatedly - the E version should by no means be a very "hard" camera to use. But given unfortunate circumstances and a sharp lens, as the Nikon official comparison shows it can indeed mess things up pretty badly...
-And I have no doubt you can get a lot worse than this.
The photographer handling the camera is supposed to know how to deal with this, that's why there's a price difference. They don't WANT idiots just looking for the latest & greatest to buy the E version, they want them to buy the normal version.
But they would never even think of releasing the E version if they did not think it was manageable - large Japanese companies are by tradition VERY careful about taking to large risks with things like this.
Now I want the D7000 E too... :-)
(well, now that Nikon HAS taken the step into the unknown, a 24MP APS camera without an AA filter shouldn't be to far out)
I expect the D5200 or whatever they're going to call it to be cheap enough for me to get one on expenses and rip it apart. Could be fun.
wayne seltzer wrote:
... I guess the software has to detect high frequency edges which are replilcated at a spacing close to the sensor pixel spacing and then fill in gaps in these sets of parallel edges? How do they remove it?
Software can blur the area and hope that the errors average to the true pixel value, or perform a wholesale replacement of pixels (i.e. synthesize pixels for you). Either way, it's a Hail Mary maneuver since the true pixel value don't exist in the image. The AA image contains true pixel values at all locations and the sharpness is True Sharpness (TM), not false impressions of sharpness.
theSuede wrote:
Yes. If you WANT to shoot at F5.6 > F9.0 all the time, use strobes and sometimes have the luck to score a red-head for the shoot - the normal version would be preferable.
If you HAVE the option to stop down just a stop more (F11?) for suspect cases you're way safe, even with the E version.
Look at the fabric - now we don't know if there's been any PP fiddling going on here, but it looks fine to me. But the fabric is pale yellow, and not really high-contrast weave patterns, so I doubt you'd get problems at F9 with this. There's a strong flat correlation in green to red>blue in that fabric, which means that moire will be less strong.
As I've said repeatedly - the E version should by no means be a very "hard" camera to use. But given unfortunate circumstances and a sharp lens, as the Nikon official comparison shows it can indeed mess things up pretty badly... http://imaging.nikon.com/lineup/dslr/d800/img/features01/img_44.jpg
-And I have no doubt you can get a lot worse than this.
The photographer handling the camera is supposed to know how to deal with this, that's why there's a price difference. They don't WANT idiots just looking for the latest & greatest to buy the E version, they want them to buy the normal version.
But they would never even think of releasing the E version if they did not think it was manageable - large Japanese companies are by tradition VERY careful about taking to large risks with things like this.
Now I want the D7000 E too... :-)
(well, now that Nikon HAS taken the step into the unknown, a 24MP APS camera without an AA filter shouldn't be to far out)
I expect the D5200 or whatever they're going to call it to be cheap enough for me to get one on expenses and rip it apart. Could be fun....Show more →
This kimono example is the exact one I was referencing (I do wonder if the D800 version above is accurately focused as it seems way too soft). Here we see a little moire with the D800E but in the full size, official NIkon samples, the same kimono is shown taken with the D800E and shows no moire - and it looks great, nothing like this smeared hair example. Thus, I guess the Nikon raw converter must have taken care of the moire pretty decently.
Even with my Sony A900, if I'm attempting to get the highest resolution out of it with a great lens, I never shoot above F8. F11 would be something I would only use if the dof was needed. I would think this would be even more the case with the D800. Thus, if I'm required to stop down past F8 only to decrease actual resolution (due to more noticeable diffraction effects) in order to avoid moire, that does not make a lot of sense to me. If it is also the case that there is no actual resolution increase with the D800E over the D800, great lenses plus correct sharpening may make more sense than dealing with moire when it occurs...but I need to see more examples.
Case 1: If you have coarse structures, i.e. a green leaf which spans several pixels, against the blue sky, then the D800E will have no problem with accurately rendering it. At the borderline between leaf and sky, the RAW conversion algorithm can make a good guess that the color of the last pixel at the leaf edge is green, and the first sky pixel is blue, even if the corresponding photosite on the Bayer pattern happens to "see" a different color. The D800 will have no problem of course, but show the borderline with slightly less contrast, due to the AA filter. A slight sharpening in post should result in an equivalent image as from the D800E. I don't think the D800 would loose any detail info here (as the structures are coarse), just contrast that is easily fixed.
Case 2: You have fine structures, close to the pixel dimension, but irregular. I.e. a landscape with some tiny spectral highlights, a grain of sand here and there, some irregular feather detail etc. The D800E will again show these details sharply, and may even have details that are missing in the D800 file, but I'm afraid it will have artifacts. I.e. the spectral highlights may have the wrong color, if they hit only one pixel. The D800 should be more accurate but less sharp. One may argue that in a print or downsized web image, the viewer may not see or notice those D800E artifacts. That's a good bet, especially in landscape shots with lots of random irregular detail. Who's gonna notice one pixel with the wrong color? But if the viewer can't see that, then I'd say he also can't see the extra detail or sharpness provided by the D800E.
Case 3: Fine regular pattern, close to the pixel dimension. This would usually be the rarest case. The D800E provides moire. The D800 will be fine.
So these thoughts make me lean towards the D800, but I'd appreciate any comments and insights. Thanks!
rico, you are a troublemaker of the best kind, lol.
'a 24MP APS camera without an AA filter shouldn't be to far out'
I doubt it's only Nikon keeping a very close watch on proceedings here...and they are to be commended for their approach in the D800.
Thanks, Philip. BTW, I appended a further sample to that FM Nikon thread. It involves some extreme moire nastiness on fabric. Qualifies as Alt, too, as the taking lens was the esteemed CZ S100/3.5 C/Y: https://www.fredmiranda.com/forum/topic/1083977/1#10330619
This whole D800 or D800E business is mind boggling. I'm curious if when shooting potential moire inducing subjects with the D800E, simply stopping down to F11, as suggested by theSuede to avoid moire, would result in the same amount of actual detail/ resolution as the standard AA filtered D800 when shot at F5.6 or F8. Even with post processing of moire, it's clear that once you get it, your screwed (the artifact pattern itself still remains or you must obliterate the detail so much as to make it blurry).
Tariq, stopping down should approximate the effect of an AA filter and, making use of an Airy disk, should be less susceptible to rectilinear image alignment versus the birefringent sandwich. As demonstrated in my test, PP with averaging algorithms cannot eliminate large-area chromatic aliasing, let alone luminance aliasing. Of course, shooting at f/11 all the time would be a handicap, even in studio environs.
carstenw wrote:
Well, no, that is too jaded. You can choose one of the following two:
1) AA filter, no or little moire, a slight softness at all times.
2) no AA filter, sharp results all the time, sometimes some moire, which may result in an effectively lower resolution after defeating it.
Which of the two you choose is probably mostly up to what you shoot, or just what you prefer. Neither is clearly better in all cases.
Dont follow this thread 100% so there might be things that I missed, but:
I'm not sure I buy 1) fully. Perhaps it's ignorance. But with sharp lenses I regularly find details travelling down to lower resolution excessively. And with that I mean that a 16-21MP file reduced to say 1.5MP (large web resolution) can produce garbled details.
It is not over sharpening I'm talking about.
And it's not moire, it's details that have "steps" in them. In one sense it's the same, but on a lesser scale locally. But globally the amount of these local details can be enormous. Depending on the subject of course.
Now, these images are captured by a cameras with AA filters. And then we (well, most of us) happily apply a downsizing algorithm such as "bicubic" (std PS) which softens edges. In fact, introducing more AA in the PP while downsizing.
Of course I'd like more control and decide myself. But I also think lack of AA filters requires a higher level of PP-knowledge than most of us here are aware about. Perhaps "higher" is the wrong word - "different" might be better.
To be very clear, this is not an argument against the D800E, we've all seen spectacular results from skilled people using the M9 that also lacks the AA filter. But I think it requires an increased level of PP understanding that I dont believe is as widespread as the idea that "AA filter sucks big time".
Wilhelm, I was only talking about the 100% view from full sized photos. If you downsize, you are introducing a lot more variables. In my experience, a single step downsize does not do photos justice.
Tariq Gibran wrote:
This whole D800 or D800E business is mind boggling. I'm curious if when shooting potential moire inducing subjects with the D800E, simply stopping down to F11, as suggested by theSuede to avoid moire, would result in the same amount of actual detail/ resolution as the standard AA filtered D800 when shot at F5.6 or F8. Even with post processing of moire, it's clear that once you get it, your screwed (the artifact pattern itself still remains or you must obliterate the detail so much as to make it blurry).
There's a very fine distinction to be made here... It's a matter of balancing your needs and average shooting situation.
IF you have the option of stopping down "enough" whenever you think there will be a need to do so - I'm guessing at F9.0 for the D800 pixel size - you get a much sharper image than if you shoot the same image with a camera WITH an AA filter at F5.6.
Reason?
Stopping down to F9+ adds a perfect amount of diffraction. "Perfect" in this case also means a uniform amount of blur over the whole image. Shooting at F5.6, with an AA filter will 1) quite certainly have less sharp edges since the lens isn't "perfect" and 2) get even worse edges by the angle effect that an AA filter has. An AA filter blurs the image MORE if the light comes in from an angle than if the light comes in "straight on".
But then you have the practical considerations - the statement above will only hold true when you have the OPTION to shoot at F9.0+. Not all lighting conditions will allow that without the use of very high ISOs. High ISOs entail noise reduction, and also less image sharpness overall due to noise dither before the interpolation stage.
I've now confirmed that we will get both versions, so some practical comparisons will be made as soon as I can find the time for it... :-)
Hmmm ... you caught my attention with the non-AA, but the fact that it is 36MP makes those pixels so dang small that the diffraction will become an issue that offsets the potential gains from the "modified" non-AA filter.
Like others have mentioned, I didn't see any real gain in detail, just contrast. Also, notice that ALL of the sample images from Nikon were shot @ f8 ... f11 or f16 would likely be introducing diffraction @ those tiny pixels rather than more detail.
I'm shooting a true non-AA in the Kodak SLR/c and the detail difference is noticeable when mated with great glass, but even with the larger pixels, diffraction can get impact things a touch (psuedo moire antidote) when stopped down far enough, even more so with the smaller pixels. I'm a fan on the non-AA, but between the combination of super tiny pixels/diffraction impact and the fact that the D800E still uses a form of AA filter (albeit modified) ... I'm not overly enthused by the non-AA filter prospects on this one (but it might still prove better than the D800 in certain applications).
AA vs. Non-AA is very real ... but I wouldn't expect a vast difference in detail between the D800 and the D800E except maybe @ wider apertures (which could be a good thing), not so much once stopped down past f8.
No emperical evidence to offer today ... just my observations / experience shooting with both AA and non-AA sensors with good glass stopped down. Will be interesting to see what good testing/comps reveal ... not holding my breath @ stopped down past f8. Not saying it isn't a kick-butt camera, just there might be a point of diminishing returns for stopped down applications.
theSuede wrote:
There's a very fine distinction to be made here... It's a matter of balancing your needs and average shooting situation.
IF you have the option of stopping down "enough" whenever you think there will be a need to do so - I'm guessing at F9.0 for the D800 pixel size - you get a much sharper image than if you shoot the same image with a camera WITH an AA filter at F5.6.
Reason?
Stopping down to F9+ adds a perfect amount of diffraction. "Perfect" in this case also means a uniform amount of blur over the whole image. Shooting at F5.6, with an AA filter will 1) quite certainly have less sharp edges since the lens isn't "perfect" and 2) get even worse edges by the angle effect that an AA filter has. An AA filter blurs the image MORE if the light comes in from an angle than if the light comes in "straight on".
But then you have the practical considerations - the statement above will only hold true when you have the OPTION to shoot at F9.0+. Not all lighting conditions will allow that without the use of very high ISOs. High ISOs entail noise reduction, and also less image sharpness overall due to noise dither before the interpolation stage.
I've now confirmed that we will get both versions, so some practical comparisons will be made as soon as I can find the time for it... :-)...Show more →
Thanks, this is exactly what I was hoping to hear. My biggest concern with the D800E is with moire in the studio or on location when photographing paintings. Since shooting this with strobes at F11 is no big deal (I usually use F8 to F11 anyway), sounds like you are saying I want run into moire issues anymore than I would with the AA filtered version. For natural landscape type subject matter, which usually consists of irregular shaped, non-repetitive subject matter, moire should not be an issue at all I guess.
mshi wrote:
Just some quick hack in PS for less than one minute, I managed to get rid of most of the moire in the sample crop.
Ah, but you had a "real" image from which to work! Without the actual fabric in your hand, or a true photographic represention, you don't know what correction (if any) is proper. Perhaps that chromatic effect is the fabric's actual appearance. The central issue is ambiguous data when sampling frequency is insufficient for the input data. It's lovely when we (or software) rejigger an image to match our expectation, but is that photography or the graphic arts? A court of law might take a dim view to photographic evidence of this kind ("Defendant wore the shimmering colored jacket, and here's a picture").
carstenw wrote:
Wilhelm, I was only talking about the 100% view from full sized photos. If you downsize, you are introducing a lot more variables. In my experience, a single step downsize does not do photos justice.
Agreed. Most actions I've seen uses two steps. For a 36MP cam perhaps three I guess. Step sizes are based on a multiple of final output size. So that means not one, but two - or possibly even three AA re-introductions.
Yes, as previously mentioned, you have not removed the artifacts created by moire beyond just the color moire effect. If you look in the gold pattern, the odd patterned artifact is still there. No way to get rid of that beyond majorly blurring it or replacing it with some guess of what the actual fabric pattern is.
That's quite a bit of color shift in the D800E "After" @ top right embroidery.
As mentioned, I'm a fan of non-AA ... BUT ... it is NOT for the faint of heart when it come to moire'. I use mine (SLR/c) with great regularity but am constantly aware of subject matter that is moire sensitive.
In that regard alone, there IS a PRICE to pay for non-AA. It is far more liberating and free flowing to shoot with my AA filter bodies without being concerned @ moire. For most applications the cost of moire' and the additional pre-capture attention / post-capture correction (when it does happen) makes it an interesting risk/reward that doesn't bode well as a 'general application' tool, imo. For me, it is my subject matter selectivity that keeps my battle with moire in check such that I rarely have to contend with it ... but I still get "bit" from time to time with it ... and it always "sucks" when it happens.
As a specialty tool (i.e. landscape/fine art/etc.), it is great ... as a general tool it is simply untrustworthy as you can't fully predict whether or not it'll show up in a given subject. I'm very glad to see that there is a non-AA option available for when my SLR/c eventually bites the dust. But, there is NO WAY that I get the non-AA option as my only/primary body for general application (read anything with people who wear clothing, etc.) ... where you don't have the luxury of time think about whether or not you should change bodies, nor the luxury of time to correct in post via PS.
Some people may reasonably associate with the non-AA filter options of medium format bodies (which I've not shot), but MF workflows (and economics) are typically very different from SLR workflows. I consider my SLR/c as an MF body with a "crop factor" that lets me use a plethora of different glass than MF itself would. To that end, I do ascribe to my SLR/c as a "poor man's MF" ... but the D800/D800E's TINY PIXELS concern me @ diffraction losses being counterproductive for stopped down applications.
I say this very strongly, and will likely be proven very wrongly by one or more talented individuals ... but for the masses, be VERY WARY of going non-AA ... it is NOT for the faint or heart or anyone wanting / expecting the ease of sooc and LR output.
If you are prepared to know that it might come with certain costs in pre- / post- capture attention, then its rewards can be pretty special ... but therein lies the "word of the day" ... it is a "special" (i.e. not general) tool that has special requirements and foibles. IMO ... most people will not do well with the special needs (attention to / correction of) a non-AA filter body. It is very alluring for the detail it can deliver, but there is a reason that there are so few non-AA filter SLR bodies options.
Again, I am a non-AA filter fan ... but, the roses do come with thorns ... and you will get "pricked" from time to time.
If you don't like getting "pricked" ... the AA filter is akin to a gardener wearing gloves.
HTH ... looking forward to seeing some EXCELLENT images from the D800E (non-moire application and f8 or faster).
BTW, I appended a further sample to that FM Nikon thread. It involves some extreme moire nastiness on fabric. Qualifies as Alt, too, as the taking lens was the esteemed CZ S100/3.5 C/Y: 