Deinterlacing HDV

[Scott Shama]

I just want to check my self...

When you deinterlace HDV, you should select upper (even) fields correct?

What is the technical problem you might have selecting lower(even)..?

I was just testing this and I can't tell a difference...

Thanks,
Scott

[Andy Wilkinson]

Scott,

I'm no expert but will try and help here! Others please correct any mistakes in my answer!

HDV is UFF, at least from my PAL Sony's it is as this is typical of PAL MPEG2. Normal PAL DV AVI files are typically (always?) LFF. If you mix the field order around and have UFF and LFF clips in the same project you can get what I call "jumpy" or "jittery" motion on any fast motion/people walking/cars driving past etc. and other issues.

Some people deinterlace the HDV's UFF by selecting Progressive (or "none" as it is termed in NLE's like Vegas) in the Project Properties so they have no field order. When the project is completed the final video file can then also be rendered out as a progressive file. In this process, the NLE's blend the fields (i.e. alternate lines) together and approximate them in a single "still image" (frame). This blending can be done in several ways (some options having advantages over others in certian situations), e.g. "blend fields" or "interpolate fields". More information and useful links in the thread I'm putting in here.

http://www.dvinfo.net/conf/showthrea...561#post822561

However, note that you effectively loose vertical resolution by doing this as the 1080i becomes 540p, if you like. Sometimes this is worth doing to remove interlaced video artifacts (horizontal "combing" or what I call a "venetian blind effect" on, typically, static "hard edges" when panning to follow a mobile subject for example.)

Also, if you crop video clips in editing you can inadvertantly swap the field order around with native interlaced video depending on whether or not the top line of your "zoomed in" clip ends up on an even or odd line.

I guess this is why the people that like the "Progressive look" (and can shoot Progressive rather than only interlaced) do so. They start out with 720p which is better than 540p and don't come up against these kind of interlaced video editing issues! At times, (often) I prefer the interlaced look as it can give smoother motion (if that movement is moderate) and the highest resolution I can attain with my current HDV gear - as long as it's not fast motion!

Everything I've written (hopefully correctly!) is from reading posts in various parts of this DVi Forum and by using the search button (using key words like de-interlacing, field order etc.) and trying it out to judge for myself. Still learning but hope this helps clarify a few things!

[Ervin Farkas]

Scott, your post is a little confusing. Deinterlaced video in neither upper nor lower field first - the point of deinterlaced video is not having fields at all. Or are you talking about which field to throw away or duplicate? You can also blend them together...

What software are you using? Different programs use slightly different terminology... What is your goal, what is your final product?

Please come back with details.

[Scott Shama]

Hi ervin,

I'm not sure what's confusing..? I'm not talking about deinterlaced video...I'm talking about deinterlacing video... The act of removing the interlacing..

Every deinterlacer I have has a setting for either upper or lower (fields) or a couple call it even or odd...

I assume this is in reference to which field is dominant in the footage you are deinterlacing...

[John Miller]

The importance of field priority is usually related to *encoding* interlaced material so that the target playback device knows which field to present first.

De-interlacing doesn't need require knowledge of field priority because both fields are being combined into a progressive frame irrespective of their original temporal priority.

If you are using a deinterlacing algorithm that requires specification of one or other of the fields then it is most likely a crude algorithm that just discards one field.

[Ervin Farkas]

Originally Posted by Scott Shama

I assume this is in reference to which field is dominant in the footage you are deinterlacing...

If by "dominant" you mean which one stays (either by itself or duplicated) and which one gets discarded, then you're right - but "dominant" can't mean "70% upper plus 30% lower field" for example. It's either one or the other (doesn't matter) or blended 50-50. Blending or duplicating one field keeps the original picture height, discarding one cuts it in half.

Best is to blend them together, but it also depends on what else you do with your footage. One common task done nowadays by videographers is to downconvert HDV to DV, in which case blending yields a smoother picture, discarding a field brings about a sharper one.

[Scott Shama]

Originally Posted by Ervin Farkas

If by "dominant" you mean which one stays (either by itself or duplicated) and which one gets discarded, then you're right - but "dominant" can't mean "70% upper plus 30% lower field" for example. It's either one or the other (doesn't matter) or blended 50-50. Blending or duplicating one field keeps the original picture height, discarding one cuts it in half.

Best is to blend them together, but it also depends on what else you do with your footage. One common task done nowadays by videographers is to downconvert HDV to DV, in which case blending yields a smoother picture, discarding a field brings about a sharper one.

It actually doesn't cut the height but the resolution.. I still end up with a video size that is 1080. Just softer...

I would love to use blend but almost all the deinterlacers I use only let you set upper or lower...

Never heard of an instance where you can select 70% and 30%. Did you mean "can" mean 70% upper etc....? Not sure what that sentence from you was going for but it didn't really make any sense..

I think you're making a little too much out of my question... Real simple... On a deinterlacing filter applied to interlaced HDV footage.. Do i select upper or lower? I've always thought the answer was upper but I wanted to know what is a negative side affect that I should see from selecting lower... From my tests, I can't see one..

Scott

[Ervin Farkas]

Originally Posted by Scott Shama

On a deinterlacing filter applied to interlaced HDV footage.. Do i select upper or lower? I've always thought the answer was upper but I wanted to know what is a negative side affect that I should see from selecting lower... From my tests, I can't see one..

Ah, OK, I think I got you now. What you are thinking of is probably the normal interlaced field order in HDV, which is upper first - irrelevant with deinterlaced video.

And yes, I went too deep into the subject - it's a complex issue. As pointed out by both replies and your own tests, it doesn't make a difference which one you choose out of the two you mention. It looks like you're using a very simple deinterlacer, with only minimal options. The method used depends on your source footage, and the result you're after... but the duplicated or discarded field doesn't matter.

[Scott Shama]

Originally Posted by Ervin Farkas

Ah, OK, I think I got you now. What you are thinking of is probably the normal interlaced field order in HDV, which is upper first - irrelevant with deinterlaced video.

And yes, I went too deep into the subject - it's a complex issue. As pointed out by both replies and your own tests, it doesn't make a difference which one you choose out of the two you mention. It looks like you're using a very simple deinterlacer, with only minimal options. The method used depends on your source footage, and the result you're after... but the duplicated or discarded field doesn't matter.

Dude..it's not irrelevant...because you are going from interlaced to deinterlaced...From what I understand you are supposed to select the correct field when deinterlacing.. I just haven't seen a visual problem when selecting lower with HDV footage so I want someone with actual technical knowledge on the subject to explain why.

..and yes it's supposed to matter which field you discard. I seem to recall reading once that it may throw sync off slightly with the accompanying audio or something along those lines... These are not simple deinterlacers either... I really wouldn't consider Magic Bullet and Nattress "simple" deinterlacers...

Also, it's not that it's terribly complex, but there are some technical aspects that I get the feeling both you and I need someone more knowledgable to weigh in on. There are two fields..you ditch one or blend the 2 to create a single image...simple... except why does it matter which field you ditch if you don't blend? I'm not getting the feeling that you maybe don't know anything more than I do about the process of deinterlacing...

Where's Graeme Nattress when you need him..?

[John Miller]

Ervin is absolutely correct.

It only matters if you are using a cheap, nasty way of deinterlacing - namely completely throwing away a field.

For any other method that uses both fields, it makes no difference whatsoever.

As I said earlier, both fields (which are separated in time by half the duration of a frame) end up at the same point in time.

Field order only matters when you are encoding to an interlaced format which, according to your description, you are not.

Graeme would say the same thing.

[Scott Shama]

Originally Posted by John Miller

Ervin is absolutely correct.

It only matters if you are using a cheap, nasty way of deinterlacing - namely completely throwing away a field.

For any other method that uses both fields, it makes no difference whatsoever.

I would love to know what a non "cheap nasty deinterlacer" is...

Scott

[John Miller]

I've already said - one that doesn't simply discard one of the fields (and therefore duplicate the other). i.e., one that has to perform some form of mathematical transformation on the contents of the individual fields to create the non-interlaced frame.

[Sean Dixon]

HDV is upper first. For any one who didn't know.

[Chris Barcellos]

Heres where I see the confusion in this post, and where I need education if I am wrong.

1. Sometimes people confuse setting up their editor properties for their timeline, versus doing the render of the time line. In Vegaus I can set up basic properties, and if my footage coming in is interlaced footage, I will select interlaced, and also for the default output files if I am going to keep them interlaced, I can select in the properties and which field to be read first. That is way I understand that.

2. Then, in Vegas, if I select progressive as a different output choice, I no longer have the option of upper or lower field first selection. (I output to a Cineform intermediate, by the way,) ]

So that leads me to believe, as others have stated here, that lower or upper field selection is only applicable if you are outputting back to another interlaced file. If I do that, I select upper field.

So, can anyone confirm that process ?

[John Miller]

Originally Posted by Chris Barcellos

So that leads me to believe, as others have stated here, that lower or upper field selection is only applicable if you are outputting back to another interlaced file. If I do that, I select upper field.

So, can anyone confirm that process ?

It doesn't matter which one you select. All upper vs. lower means is which one comes first in time.

Assume you record at 50i (to make the numbers easy). Every 1/50 sec, the camcorder records one field. Since a frame comprises two fields every 1/25 sec, you need a way to specify which field was first in order to recreate the fields from the frame - i.e., are all the odd numbered lines the first field or are all the even numbered lines?

The recorded field sequence may be:

F1 F2 F1 F2 F1 F2 F1 F2 F1 F2 F1 F2

You can make a frame-based video from single fields simply by selecting all the F1's or all the F2's.

The only difference it makes is that the video is shifted by 1/50 sec.