5D Moire Work-around

commanderspike

commanderspike

Carbonite Member
Here is my fix for moire.

No software / FCP stuff done in post to this, or digital filters necessary, it's all done in-camera. I didn't use any filters on the lens itself. The footage is straight from the camera with a Zeiss lens.

5dmoirefixed.jpg


This is a frame grab of the original H.264 footage at 100% and it was shot at ISO 3200. Quite clean considering that. Viewed on Vimeo or a TV it would look even better, you wouldn't notice anything unusual.

I'll post more updates soon along with more details about how it works.
 
Last edited:
Hopefully the horrendous amount of sharpening of the one on the right is not a requirement for the method!
 
Looks to me like going to F8 - F16, raising the iso to 3200 and focusing just a bit past you intended target ( the chair ) making it seem in focus but really its just slightly out ?

hopefully its more magical then that... :smile:
 
Here is how I achieved the look:

- No-name diffusion filter I found on eBay for £5
- Close down the aperture (F5.6 on this occasion)
- No back focussing required, works at any ISO
- Sharpness plus 7 in-camera rather than afterwards in software

ISO can be anything. I had it at ISO 3200 as I had to have the lens at F5.6 under dim interior light. But it will work at ISO 100 too, in good light.

There isn't an enormous amount of moire in the right hand picture. That is how the eye sees that chair. The moire is completely removed. The fabric of the chair looks wavy because it isn't flat, it's like a rippled cloth so some of the small holes look closed-up and darker from far back.

Moire is all coloured, and it moves and shimmers between frames on the slightest camera movement. Very distracting.

I used picture profile Faithful. The image is being downsampled from a very high resolution 22MP still, which is what causes the moire. But since the original image has so much detail in it, if you boost the sharpness in-camera to +7 rather than afterwards in software this causes the downsampling to retain a lot of the fine detail even with the diffusion filter. The benefit of doing it like that is that you can then SOFTEN the whole image in Final Cut Pro to smooth out any unwanted hard edges while retaining the detail you need. The key thing is, that due to the way the 5D handles the in-camera sharpness levels the original detail is intact in the source footage.

The diffusion filter simply acts as a stronger AA filter over the sensor.

The image doesn't lie, what you see is what you get. When moire crops up all you have to do is this, and then soften the image a little bit in post so that it looks consistent with the shots made with sharpness levels in-camera at the normal level.

I really like the fixed footage. It doesn't look over-sharpened, remember that shot above is at 1:1 100% magnification in FCP.

PS don't buy a diffusion filter until I figure out what the hell this freaky thing is I got on eBay!! They come in lots of different types. This one has personality but no label :)
 
Last edited:
I've not yet had chance to test it's affect on aliasing fully Ian, but will do soon. It should do, as the diffusion filter is simply a strong AA filter.

The 5D and 7D, and now 1D Mk IV all lack a strong AA filter necessary for video, because in adding one they'd hurt the stills performance. People like to squeeze out all the sharpness from their expensive Canon L or Leica lenses I guess and with good reason, but the 22MP footage doesn't NEED to be so sharp when it's being downsampled to 1080p. That just causes us video guys moire headaches :)
 
Good find, Commanderspike! Please do post your results on aliasing as well as the brand of the diffuser.

Any 7D user willing to give this method a go? Maybe even face the infamous 720p mode?
 
Last edited:
can't over sharpening in camera lead to other sorts of artifacting other than hard edging? or am i WAY off on that?
 
It can exaggerate artefacts already in the image, but since the diffusion filter removes those in the first place - the sensor doesn't output anything too challenging to the image processor.

It's a more filmic look. Video cameras should have a strong AA (anti-aliasing) filter.
 
Video cameras should have an appropriate AA filter, yes. One that filters out fine detail that's beyond the sensor's capability to properly sample.

The danger of employing something like this on a camera like these DSLRs is: the native resolving power of the video portion of the cameras is very low. As in, about standard-definition, or maybe, at most, near to 720p. So the only reason they look sharp at all is because of the aliased detail that's getting through; it's a "shortcut" or "cheat" to making sharp images. If you start taking that out, you'll leave the cameras revealed for what they are: medium-def.

Commander's technique here is to filter out excess detail, and then attempt to compensate by digital sharpening. Filtering out excess detail is, basically, the only way to combat aliasing. But it's trading one artificial detail-enhancement technique for another, trading aliasing for edge enhancement.

Commander, your first post is confusing -- it says "I didn't use any filters on the lens itself", but then in post 6 it says "I used a no-name diffusion filter". Are you saying that you somehow wedged this no-name filter behind the lens, in the mirror box?
 
commanderspike said:
There isn't an enormous amount of moire in the right hand picture. That is how the eye sees that chair. The moire is completely removed. The fabric of the chair looks wavy because it isn't flat, it's like a rippled cloth so some of the small holes look closed-up and darker from far back.
Click to expand...

Ah ! I see, sorry my mistake.

Great method, seems to work well.
 
Barry_Green said:
Commander, your first post is confusing -- it says "I didn't use any filters on the lens itself", but then in post 6 it says "I used a no-name diffusion filter". Are you saying that you somehow wedged this no-name filter behind the lens, in the mirror box?
Click to expand...


Ya he did say no filters, and well... that is a filter. I messed around with diffusion filter this summer with only mixed results and I wasn't happy with the overall effect it had on the footage.
This is an age old method, but as always, there are many ways to cook a goose so Commander... can you post some footage ??
 
I think after re-reading that it's pretty clear that he was indeed putting the filter on the back of the lens, not the front.
 
I had the diffusion filter in the mirror box behind the back of the lens, but this has lead to a few issues. I'm gonna test more, but it looks like I may have to try out a few more diffusion filters and techniques - worked well for the chair but it doesn't work well enough in all situations where moire crops up especially the more subtle occurrences. The diffusion filter is quite a poor substitute for a strong AA filter over the sensor really.

Maybe I'll be better stuffing a load of lingerie in the mirror box. What do you think? :)
 
commanderspike said:
Maybe I'll be better stuffing a load of lingerie in the mirror box. What do you think? :)
Click to expand...

I tried this already, it can work though you have to find the right stocking to do this. Avoid the naked color stockings because they just they lower contrast too much and give a tan wash to the image. It's tough to hold it back there and I worried about strands getting loose.
 
ASG said:
Hopefully the horrendous amount of sharpening of the one on the right is not a requirement for the method!
Click to expand...


Maybe you are seeing something I'm not. Looks a lot better than the left side. If you have a cleaner solution I'm waiting.
 
Barry_Green said:
Video cameras should have an appropriate AA filter, yes. One that filters out fine detail that's beyond the sensor's capability to properly sample.

The danger of employing something like this on a camera like these DSLRs is: the native resolving power of the video portion of the cameras is very low. As in, about standard-definition, or maybe, at most, near to 720p. So the only reason they look sharp at all is because of the aliased detail that's getting through; it's a "shortcut" or "cheat" to making sharp images. If you start taking that out, you'll leave the cameras revealed for what they are: medium-def.

Commander's technique here is to filter out excess detail, and then attempt to compensate by digital sharpening. Filtering out excess detail is, basically, the only way to combat aliasing. But it's trading one artificial detail-enhancement technique for another, trading aliasing for edge enhancement.

Commander, your first post is confusing -- it says "I didn't use any filters on the lens itself", but then in post 6 it says "I used a no-name diffusion filter". Are you saying that you somehow wedged this no-name filter behind the lens, in the mirror box?
Click to expand...

Has anyone tested SoftFX to see what happens if anything?
http://www.tiffen.com/sfxpics.htm
 
He's referring to the digital sharpening "halos" that are surrounding the image. Look along the top of the back of the chair, where the brown of the chair meets the yellow of the wall, and on the photo on the right you'll see there's a big ol' white halo there, on top of a thick dark black line that's been drawn on the top of the chair. That's the effects of electronic "sharpening".
 
You must log in or register to reply here.
Back
Top