Why don't we need an anti-alising/low-pass filter in front of the sensor anymore?

donaldejose

Good explanation and one good reason to welcome the higher megapixel sensors. Their feared low IQ at high ISO has not proven to be true.

MikeGunter

Hi all,

Thanks Ade.

@aquarian_light Sorry, I thought you understood that the real quandary about moire was in video, hence the two versions of D800 in the first place. I so sorry for the confusion.

Moire is quite real, likely not noticeable if you don't do any video, not something that you can notice, if you don't do video.

My apologies and by best,

Mike

donaldejose

It is not a confusion Mike. It is a good question and a good point. Just seems very few here do much video so it is an issue we under report on.

Ade

Mike, I believe in real-world testing the video moire difference between D800 and D800E is slight to negligible depending on the circumstance. That is to say, they both can produce video moire in abundance. The reason is because the AA-filter in the D800 is already very weak (even for stills) and as you know a lot of the moire is produced by line skipping, etc., which is identical in both models.

Since the D800/E has to downsize the image to video so much (by a factor of 18), it really needs a very strong AA filter for video -- one that would not be acceptable for stills. The only real solution is to get an external AA filter like the ones from Mosaic Engineering.

It will be interesting to see how the D7100 will perform. Technically speaking, video moire produced by downsizing is a tradeoff. Nikon could choose to tune its image downsizing algorithm to reduce moire but at the expense of sharpness. Or preserve sharpness at the expense of moire.

E.g., newer Canon DSLRs are more resistant to moire, but produce much softer video -- not necessarily because of the strength of their AA filters but because of image processing choices.

Elvishefer

Could it have been taken off the 12mp D700?

I shot a kitchen once and the 'spotted grid' on the oven/range glass door, however it's manufactured, set off moire so bad it was shocking. I don't shoot a lot of kitchens so realistically the filter could be removed for the other 99.whatever% shots I take.

I never felt the filter saved me from moire after that, though.

heartyfisher

@ aquarian_light : Not disagreeing with your conclusions and opinions :-) but this pixel peeper just wants to correct some "facts" in your posts :-)
1) The bayer array is made up of 4 pixels not 3 ie 1 blue 1 red and 2 greens its a 2x2 matrix or 4 pixels.
2) the demosaic process takes the 4 different colored pixel and creates 4 pixels not 1.
3) I doubt that it is "exceedingly difficult" to create an AA filter for higher MP sensors cos the cheaper D5200 and the D800 has one.
4) The D800e actually has an AA filter just one with a much reduced effect. ie. it splits then recombines http://imaging.nikon.com/lineup/dslr/d800/features01.htm
5) Moire can be seen fairly easily if you want to. good ol' KR shows how here .. http://kenrockwell.com/nikon/d800/vs-d800e.htm

@ MikeGunter.. there seems to be an AA filter you can buy to remove video moire on the D800/D800E http://www.mosaicengineering.com/products/vaf/d800.html

aquarian_light

3) I doubt that it is "exceedingly difficult" to create an AA filter for higher MP sensors cos the cheaper D5200 and the D800 has one.

5) Moire can be seen fairly easily if you want to. good ol' KR shows how here .. http://kenrockwell.com/nikon/d800/vs-d800e.htm

yes, but both the 5200 and the 800 have AA filters that are definitely on the weak side compared to others.

true, but notice both cameras produced moiré to some extent, and even KR notes to 800E is sharper

Reese

My guess is that if the D800e users aren't complaining about moire then the D7100 users won't as well considering that it has a sensor that is even denser than the D800e's sensor.

heartyfisher

@ donaldejose : You mentioned you posed this question on a different forum.. can you provide a link to it? Just curious what others think on this issue.

@ Reese : Welcome ! first post huh!

I was looking at mosaicengineering's AA filter for the D800.. just thinking about it .. does that mean that the filter is optimised for 1024p images on a FX sensor ? doesnt that mean that with that AA filter on you would get unsharp images for the DX and CX format video? and for stills it would be even worse ! in effect making all images 2MP in resolution? right ?

donaldejose

heartyfisher: here is the link you requested. http://www.nikoncafe.com/vforums/showthread.php?t=355465

TaoTeJared

@MikeGunter - if you let me know what type of subject you would like to see if Moire shows I'll do my best to take it with my D800 - Fair warning though - I stink at video so it will be "raw".

I'm seriously just sitting back and laughing at the comments trying to prove a theoretical something that absolutely fails in real world practice because they doesn't actually understand what causes it and even a much lesser understanding on how Nikon tries to fix it. No 1080p video moire because it is compressed? You are being sarcastic on that right?
Nikon has never published anything on how their software side deals with Moire. Since it is not spots on sensors, maybe they will some day - but that hasn't come yet.

Ade

I was invited to shoot a fashion show in Toronto this past weekend. My new D800E came literally one day before the event. Runway shoots are always quite demanding: should I bring the untested D800E with possibility of moire (lots of fine fabrics) and "slow" 4 fps, or take my trusty and faster D700?

I made a few test shots to check focus and it was a no brainer decision: D800E primary, D700 as second BTS camera since I will have "backstage" access.

Here's an example close-up from the shoot:

(Red dress crop)

I picked this one as an example because of the reds. My D700 would have blown all the reds and made a complete mess out of the dress. If you can view the above on a wide-gamut monitor (or at least one that doesn't clip the reds) you can see that the details are preserved.

The hardest part about runway shoots is focus. The models are constantly in motion (even when they briefly "pose" at the end of the runway). More often than not you'll get some softness either due to motion blur or because the AF didn't focus exactly:


(Extreme crop)

But with the D800E I can still capture a ton of details. 36mp worth of details. You can still see fine texture in her hair, it didn't resolve into a black blob. When I reduce this picture for printing (or to a D700 equivalent 12mp) the final image looks superb.

More importantly, with over a dozen models, countless outfit changes, and hundreds of shots, there was exactly ZERO moire. I'm glad I opted for the superb quality of the D800E instead of the "safety" of my D700's AA filter.

-Ade

Ade

Nikon has never published anything on how their software side deals with Moire. Since it is not spots on sensors, maybe they will some day - but that hasn't come yet.

http://www.nikon.com/about/technology/rd/core/image/image_processing_e/index.htm

"Our proprietary image processing engine makes it possible to use an OLPF with the birefringent index set relatively lower, to increase resolution, and still suppress false color." -Nikon

(As discussed in the Coolpix A thread last week)

-Ade

haroldp

" Most high-end fashion and commercial photography for the past 5+ years have been shot on relatively low-density MF digital backs with no AA filters. Maybe the only exceptions are people who are still shooting film. "

MF backs are up to 80 mpx and at 36 X 48 mm exactly twice the area of an FX sensor, meaning slightly smaller pixel pitch than a D800.

MF lenses having to cover a wider image circle, do not have the resolution at the focal plane of the best 35mm lenses, even though total resolution is better because magnification of the captured image for any given display size is less. The lower resolution at about the same pixel pitch reduces artifacts.

In practice, I have been usin a Leica M9 since it came out. At 18 mpx it's pixels are larger, and Leica lenses are at least as sharp as Nikon's best, and have had no moire problem that I noticed.

... H

Ade

MF backs are up to 80 mpx and at 36 X 48 mm exactly twice the area of an FX sensor, meaning slightly smaller pixel pitch than a D800.

No. Even the IQ280 has a pixel pitch of 5.2 microns. D800 has a smaller pitch of 4.88 microns.

Besides, 80mp backs are relatively recent. The P25 "only" had 22mp and covered 645 with its whopping 9 micron pixel pitch. Studios around the world bought them en masse. For $30,000 when it came out, it didn't come with an AA filter.

MikeGunter

@MikeGunter - if you let me know what type of subject you would like to see if Moire shows I'll do my best to take it with my D800 - Fair warning though - I stink at video so it will be "raw".

Hi TTJ,

There is a posting referring to the filter that Mosaic Engineering makes (thanks heartyfisher - I've been aware of them and their filter, I'm just not too keen on the fix, that is to put the filter on/off for video/stills as it is a kulgey fix, IMHO), but if you're still game, TTJ, shooting a brick wall with a wide-ish lens or a staircase or any parallel set of objects should do.

In video, what moire tends to do is 'pulse'. One can 'engineer' pulsing by encoding video badly.

In the video on the thread located (and I'm not sure how the links show up, so I'm posting somewhat blindly, but they are the YouTube 'D7100' video links, please forgive. .

http://www.forum.nikonrumors.com/discussion/690/video-from-d7100-on-youtube#Item_3

The video the video with the length of 2:17 shows moire and a lot of other, well, crap.

First, I really don't think the it's D7100, or if it is, the owners know little about encoding video and aren't doing a service to Nikon.

Secondly, interlaced video, which has a place, will always have moire in a progressive space unless deinterlaced in the encoding process.

Finally, video and still are vastly different, with complex color space and color definitions. The Panasonic GH3 encodes at 95Mbps and will encode and write video until the card is full - while all the Nikon models (at best - up to the latest iteration it was 20 minutes) have a 30 minute maximum, and even that is further reduced by complexity of content. The bit rate on the card is a maximum of 24Mbps. It does have uncompressed out at the HDMI port, but geez, come on, a dinky $1300 micro 4/3ds is still out-pacing a king in this area (my GH2 is still a better movie camera - and there it is - and yes, I have a Nikon adapter).

Again, I doubt in the long run, that moire will have much impact on stills, but in video, it could have some impact, and on stills, if display on displays and monitors, it could.

My best,

Mike

SquamishPhoto

I've been commissioned to be the DP for a film short to be shot this spring and I will be using the D800 exclusively for the film. If I run into this issue while shooting I will make a post here with any relevant findings.

heartyfisher

@ade : Re the 2 photos you posted, how were they post processed ?

heartyfisher

Nikon has never published anything on how their software side deals with Moire. Since it is not spots on sensors, maybe they will some day - but that hasn't come yet.

http://www.nikon.com/about/technology/rd/core/image/image_processing_e/index.htm

"Our proprietary image processing engine makes it possible to use an OLPF with the birefringent index set relatively lower, to increase resolution, and still suppress false color." -Nikon

(As discussed in the Coolpix A thread last week)

-Ade

from that article "Sophisticated algorithms faithfully interpolate color that the captured image lacks, using color information gathered from adjacent and even further-away image pixels. Unless these algorithms are written properly, "color aliasing (false coloration)" not in the original scene will be recorded in the image, degrading the verisimilitude of the image to reality."

It looks like they may have indeed included the "Matrix" algorithms I mentioned earlier. That would explain the lower occurance of moire and higher iso capability of these high MP cameras.

heartyfisher

Video showing moire on D800..

PB_PM

Looks like the results from a review I saw shortly after the D800 came out. Moire is a big problem in video. Not with all subject matter, as seen in that video, but it is there.

Ade

@ade : Re the 2 photos you posted, how were they post processed ?

After the shoot, I used Aperture 3 to vignette, set color temperature, then add very slight edge sharpening and contrast to all the pictures. For the 2nd picture above I turned off edge sharpening, to show the slight blur visible at that cropping level. I'm still experimenting with the amount of sharpening to use with the D800.

On my Macbook Air, Aperture really struggles with the 36mp files, I think because the laptop only has 4GB of RAM and a relatively weak GPU.

TaoTeJared

http://www.nikon.com/about/technology/rd/core/image/image_processing_e/index.htm

"Our proprietary image processing engine makes it possible to use an OLPF with the birefringent index set relatively lower, to increase resolution, and still suppress false color." -Nikon

from that article "Sophisticated algorithms faithfully interpolate color that the captured image lacks, using color information gathered from adjacent and even further-away image pixels. Unless these algorithms are written properly, "color aliasing (false coloration)" not in the original scene will be recorded in the image, degrading the verisimilitude of the image to reality."

It looks like they may have indeed included the "Matrix" algorithms I mentioned earlier. That would explain the lower occurance of moire and higher iso capability of these high MP cameras.

Sorry Ade that article says absolutely nothing - besides it is for the D100 in 2003. Every manufacture that uses Bayer sensors has to use matrix algorithms to get color - that is what the Bayer sensor technology is.


@Mike - I'll do my best and think I have a good spot - should be nice here again, 60's and the last snow is almost gone!

heartyfisher

@ade : Re the 2 photos you posted, how were they post processed ?

After the shoot, I used Aperture 3 to vignette, set color temperature, then add very slight edge sharpening and contrast to all the pictures. For the 2nd picture above I turned off edge sharpening, to show the slight blur visible at that cropping level. I'm still experimenting with the amount of sharpening to use with the D800.

On my Macbook Air, Aperture really struggles with the 36mp files, I think because the laptop only has 4GB of RAM and a relatively weak GPU.

Thanks, I like the "colour depths" ( not sure how else to describe it :-) ) though I do see a very very slight yellow shift.. but it adds a nice warm feel. So tempted to get a D800 .. and forget about the D400 :-( (opps wrong thread ! ) :-)

Ade

Yes it was a mixed lighting scenario and as can be seen in the first photo there are warm yellow room lights right by the runway, which wasn't ideal. I was able to set up one strobe past the end of the runway (1x Profoto AcuteB2 w/ a 45" umbrella). The color balance is neutral (set using a ColorChecker) but depending on where the model was in relation to the room lights there might be some additional yellows falling on their faces, etc.