Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Any advice for removing aliasing in post?

Matt W.

Well-known member
Joined
Oct 3, 2009
Messages
1,350
Reaction score
1
Points
38
From h264? I might have used a cheapo camera that should not be named here, but the advice on this forum is generally better than elsewhere. Just mask the offending area and blur it or apply noise reduction? What's the best technique? Thankfully the worst parts aren't visually significant within the scene; they just attract attention by virtue of flickering so much.
 
From h264? I might have used a cheapo camera that should not be named here, but the advice on this forum is generally better than elsewhere. Just mask the offending area and blur it or apply noise reduction? What's the best technique? Thankfully the worst parts aren't visually significant within the scene; they just attract attention by virtue of flickering so much.

You need to use tunable frequency cutoff filter, where you can adjust it, depending on the material. FilmMaster has just such filter and it's name is ... Alias filter:-) I'm sure there are many other tools, that do same thing...
 
You need to use tunable frequency cutoff filter, where you can adjust it, depending on the material. FilmMaster has just such filter and it's name is ... Alias filter:-) I'm sure there are many other tools, that do same thing...

Caveat: I haven't tried this, but it would probably work.

Do a Fast Fourier Transform on the image, blur the transformed image, and then apply an inverse FFT. Apply that as a patch over the offending area.

That should do what Jake says - apply a tunable frequency cutoff.
 
That's a bit beyond me. Any way to do this just using After Effects plug ins? I'm glad to learn it's possible, at least.
 
The possibility all depends on the strength of the aliasing and to what extent the extra high frequency content was folded back into the lower image frequencies. It can be, on strong aliasing that the content folds back enough to low frequencies while retaining enough contrast that even a very strong blur will not remove the damage.

Graeme
 
Thanks, I can see that happening in some cases. The aliasing is really high contrast and kind of weird looking, but thankfully I avoided it most of the time. To all those with aliased wide shots, throw your shot slightly out of focus when zoomed in to 10X or whatever. If you're pulling focus, set your focus mark for the background based on the closest high-frequency brick building being slightly out of focus in 5x or 10x mode. This really works but you have to go pretty far out of focus (not too big an issue since the cameras aren't very high-res).

Btw, the more I use these, the less I want to for larger projects. I'm glad Red has quit the consumer sphere since these will own it (they're cheap and easy to operate), but they're totally inappropriate for any production that can afford to rent better. I take back anything I said comparing these with what Red's doing. Completely different cameras for completely different markets.
 
Last edited:
With relatively little effort I had a lot of luck. That doesn't happen often! Here are some before-and-afters (at 200%, as PNG files):
 
Last edited:
Thanks, I can see that happening in some cases. The aliasing is really high contrast and kind of weird looking, but thankfully I avoided it most of the time. To all those with aliased wide shots, throw your shot slightly out of focus when zoomed in to 10X or whatever. If you're pulling focus, set your focus mark for the background based on the closest high-frequency brick building being slightly out of focus in 5x or 10x mode. This really works but you have to go pretty far out of focus (not too big an issue since the cameras aren't very high-res).

Btw, the more I use these, the less I want to for larger projects. I'm glad Red has quit the consumer sphere since these will own it (they're cheap and easy to operate), but they're totally inappropriate for any production that can afford to rent better. I take back anything I said comparing these with what Red's doing. Completely different cameras for completely different markets.

I fell into that trap - Watch Reverie, get excited, then actually use the damned thing and vow, never again.
 
It's not so bad if you treat it like a nice toy rather than a bad serious camera. But for anything other than youtube video type stuff, b camera, no crew short films or what have you, it's a silk purse/sow's ear situation... I'm just glad to have found some way to deal with aliasing effectively, not that it's necessarily worth the excessive effort bringing things into After Effects.
 
It's a bit hit-or-miss (though so far fine for all the shots I've needed to fix), but I basically just did careful noise reduction and blurred the color channel.

I've been using neatvideo for After Effects (recommended), for which I sampled a gray fabric exhibiting rainbow aliasing as the sample noise. I then set the chroma NR and low-frequency NR up pretty high, leaving the luma-based NR settings a little lower. I turned off adaptive filtration and set the temporal window to 5 frames, which is the key.

I then duplicated this layer, placed the duplicate layer on top of the original, and applied an additional 12-pixel radius fast blur and set the transfer mode of this layer to "color." This removed a lot of still-visible color aliasing, but at the cost of perceived saturation and color detail.

In some cases this made the footage rather mushy, so I masked out the offending areas and left the rest as-is. Unfortunately this is slow and kind of requires camera movement, even if just a little, to work, but I'm not concerned since I don't get aliasing on static shots. I always set the shot slightly out of focus.

Some more examples:

before3.png

Before.

after3.png

After.

before4.png

Before.

after4.png

After.

The first one was sample footage from dpreview.com, which had tons of aliasing and almost kept me from buying the camera (except I largely wanted it as a still camera so if I can use it for video that's just gravy), the second is cloth pushed 3 stops in after effects. The fabric pattern doesn't get too decimated. I know this isn't an elegant way to deal with the problem but it's a big help for me.

If you have any terrible examples I'd be willing to use this a try on it.
 
Last edited:
Caveat: I haven't tried this, but it would probably work.

Do a Fast Fourier Transform on the image, blur the transformed image, and then apply an inverse FFT. Apply that as a patch over the offending area.

That definitely wouldn't high-pass the image. Convolving (filtering) an image is equivalent to pixel multiplying its Fourier transform. Actually (I played around with this in Matlab), I think all this does is apply a reverse vignette (ie darken the center of the image). Might this be because the reverse of the above rule is also true (convolving FTs equivalent to multiplying)?

If you wanted to filter the high frequencies, you would take the FFT, and then matte out a circle and IFFT. But the results of this are not very different than a Gaussian blur. In fact if you were to pixel multiply the FFT by a Gaussian distribution and IFFT it would be the the same as a Gaussian blur (because the FT of a Gaussian is a Gaussian).
 
Back
Top