Video Aliasing

Jon Rosenbloom · February 5, 2005

I bet this has been ansered a million times, but, where does video aliasing come from? I just watched a new dvd of a short film I shot a while back on the SDX-900; it is rife w/ aliasing - mainly along vertical and "hot" edges, that throughout the rest of post production was not evident. My guess is this has something to do w/ Final Cut, or IDVD, since I cut some of the same footage into my reel (from mini-DV dubs) on an Avid and don't see any aliasing there.

DavidSloan · February 5, 2005

Alaising is a product of interlaced video. I don't know anything about the SDX900, I suspect it's not a 24P camera...so there is nothing you can do about it. If the material is originally film or 24P and viewed on a progressive scan DVD player then the problem is solved.

Phil Rhodes · February 5, 2005

Hi,

> Alaising is a product of interlaced video.

No, it isn't, it's a term used to describe stair-stepping in analoge-to-digital conversions, including imaging and sound.

> I don't know anything about the SDX900, I suspect it's not a 24P camera...

It most certainly is a 24p (or 25p) camera.

> so there is nothing you can do about it. If the material is originally film or 24P and

> viewed on a progressive scan DVD player then the problem is solved.

I suspect what our correspondent is seeing is either edge enhancement or very harsh hilghlight clipping; I'd have to see it. In any case, the playback and display devices won't solve the problem.

Alternatively he might be seeing interline shimmer, which is a product of interlacing, but which is often also an artifact of an interlaced display device irrespective of the capture medium.

Need more info.

Phil

DavidSloan · February 5, 2005

Thanks for the correction, Phil. Just curious but is interlacing and stairstepping related?

February 5, 2005

Hi,

> Alaising is a product of interlaced video.

No, it isn't, it's a term used to describe stair-stepping in analoge-to-digital conversions, including imaging and sound.

Phil

<{POST_SNAPBACK}>

It's not only true for A/D conversion. Everything that is "sampled" with too low a resolution can cause "aliasing". This can also be a digital source of higher (or "infinite") resolution / frequency (i.e. 3D renderings).

Maybe it's the mpeg2 DVD compression?

-k

Phil Rhodes · February 5, 2005

Hi,

> interlacing and stairstepping related?

Not intrinsically, although most digital image capture systems sample one row of pixels for every row of the interlace (if any), so there will be aliasing along with the line shimmer.

In either case, I'm not sure it's a big deal. You can't even nearly see HD pixels on a filmout - they're buried in the noise. Sorry, grain, pardon me. When the world goes universally to 4K it'll be a complete non-issue for theatrical features, as it should be for any kind of properly-filtered imaging.

Phil

Jon Rosenbloom · February 6, 2005

So you need more info? What I'm calling "aliasing" might be what you call stairstepping. In narrow vertical objects - like unistrut columns (the film takes place in a big warehouse store) - if the object has any kind of an edge light on it, the edge will flicker w/ any kind of camera movement. There's a shot where a woman walks past a row of freezers - all glass and aluminum doors w/ vertical flourescent bulbs inside, and the flo's - on the dvd the director gave me - display a lot of flicker. Now, the actors often have a little edge light on their arms or shoulders, but those edges don't flicker, or as I say "alias." The condition is limited to really narrow objects. Anyway, we shot project in 24p.

But, now that I've brought it up ... I still don't get the whole video capture thing. If you shoot in a progressive mode, the tape that you pull from the camera is still interlaced. Why does it look any better than if you just shoot in old-fashioned interlace format? Furthermore, if I take the 4:2:2 footage from the SDX-900 and dub it to mini-dv, aren't I now working w/ 4:1:1 footage. How is it that this footage still looks "better" than stuff I shoot on the DVX-100? (Obviously, the same questions apply to film to beta work.)

Phil Rhodes · February 6, 2005

Hi,

First off you need to look at something other than a DVD. There are so many variables intrinsic to the encoding process that all bets are pretty much off.

> I still don't get the whole video capture thing. If you shoot in a progressive mode, the

> tape that you pull from the camera is still interlaced.

No, it isn't, it's in some kind of 3:2 pulldown, which presumably you removed from it before editing.

> Why does it look any better than

> if you just shoot in old-fashioned interlace format?

I'm not going to go into this here as it's been discussed ad nauseam before, use the search function, but really - shouldn't you have found all this out before shooting the format?

> Furthermore, if I take the 4:2:2 footage from the SDX-900 and dub it

> to mini-dv, aren't I now working w/ 4:1:1 footage.

Yes.

> How is it that this footage still looks "better" than stuff I shoot on the

> DVX-100?

Argh. This really irritates me. You know when you read trade publications, you find people looking down on cameras like the PD-150 and DVX-100? Well that's why. To answer your question: because the fifty-grand SDX-900 is a much, much, much better camera than the five-grand DVX-100. It has better lenses, better sensors, and better image-processing electronics. This is the difference between 800-line broadcast cameras and 500-line consumer cameras. You noticed - well done!

Don't make the mistake of comparing tape formats and assuming any camera with such-and-such a recorder welded on the back end will be equivalent. This is an incredibly common mistake. A PD-150 records DVCAM; so does a DSR-570, but the DSR produces vastly better images.

Phil

Rob Belics · February 6, 2005

Aliasing is caused by undersampling a scene. If you had an edge of an object, say a table edge, and your camera scanned that edge but the sensors didn't line up precisely with it then you would sample above and below the edge itself.

When displayed on a screen that edge would either have to be "quantized" into a single point using information scanned above and below the edge, or, if your display matched the camera pixel for pixel, then at some point you would have to display the pixel above, then below, the actual edge which would cause "staircasing" because an "alias" for the actual point is created.

I should say that every scene is undersampled. There is no perfect sensor that can capture every point of light.

Edited February 6, 2005 by Rob Belics

Jon Rosenbloom · February 6, 2005

> Why does it look any better than
> if you just shoot in old-fashioned interlace format?

I'm not going to go into this here as it's been discussed ad nauseam before, use the search function, but really - shouldn't you have found all this out before shooting the format?

Phil

<{POST_SNAPBACK}>

I've tried the search engine on this board a bunch of times, but it never works very well. In fact, I used "aliasing" as my search term, and the engine returned a grand total of one post.

As for my curiosity about the technology, it's just that: curiosity. It's one thing to know that mini-dv is pretty good (I've had great results w/ the dvx-100a), dvc-pro is better, and hd is "really good," but I'd like to have a better handle on the how & why of each format.

Anyway, I'm going to write off the aliasing as a result of looking at a dvd that someone burned. It's really not there in the vhs of my reel.

Thanks, Rob, for your concise, and patient, explanation.

Last, I'm told the film in question will be available at www.cinequestonline.org from Feb. 14 - 28, under the title "Superstore." If you watch it all the way to the credits, then I guess my anonymity will be blown; but that's a risk I'll live w/ :rolleyes:

Michel Hafner · February 8, 2005

I should say that every scene is undersampled. There is no perfect sensor that can capture every point of light.

<{POST_SNAPBACK}>

It does not have to since the lense already bandlimits what reaches the sensor. And if that's not enough there are anti aliasing filters required. It's not an issue with careful system design.

John Sprung · February 8, 2005

Aliasing is a huge and fascinating subject. What you need is a good introductory book on sampling theory. Alas, I don't know of one to recommend. You might google around for Harry Nyquist and Claude Shannon, who did the fundamental work.

-- J.S.