PDA

View Full Version : Progressive Video Questions



Ryan
05-28-2003, 01:13 PM
I am shooting a lot more video for the Internet these days and I am a bit confused about how the toaster deals with progressive video.

I have been shooting with a Canon XL1s in frame mode (progressive) but the toaster still dose the Transitions, Effects, etc. in interlaced mode. If I render it out as progressive from the toaster it seems to do field blending in order to deinterlace. Logically that would blur my source video if it blends the fields.

Is there a way to tell the toaster that the source isn't interlaced and to only deinterlace the effects? Or would I be better of to deinterlace with virtual dub when I'm done? Or is the toaster taking care of things under the hood and I just didn't know it.

I want to get the best quality I can out of these videos. Internet video is bad enough without messing things up before its compressed.

Please share any tips/tricks related to Internet video you may know too.


Ryan

Admitting Ignorance is Prerequisite for Gaining Knowledge.

vanguard
06-02-2003, 07:59 AM
My .02 would be to forget the hasssle of progressive scan.

Interlaced works fine and like you say internet video is pretty crappy.

I doubt anyone would be able to tell (via the web) whether you shot on Progressive or not.

Now in a home theater environment.....

If what you are shooting is for web use only, you are squeezing 55 gallons of video through a 4mm opening...

SVHS and D-5 are practically impossible to tell apart at web transmittable resolutions...

Ryan
06-02-2003, 10:51 AM
I understand that its overkill, Part of this is that I have settings in the toaster that I dont understand and that bothers me. I am also do VFX projects on my own time. So It would be nice to understand whats going on under the hood so I can use it later.

Web compression is terrible. But we do release alot of content at 700k per sec. The picture is suprisingly clear at that bitrate. And the bit rates and res are going to keep going up.


Ryan

Admitting Ignorance is Prerequisite for Gaining Knowledge.

mblade
06-03-2003, 02:42 AM
I agree with Ryan, it bugs me to not fully understand an option in a piece of software!

As a result I have spent many sleepless nights pondering the implications of progressive video. Here are my naive views. Please correct me if I'm wrong...

Traditional interlaced video works by recording fields at twice the frame rate. So for pal we capture 50 fields per second with each field captured at 1/50th of a second.

The shortfall here is that deinterlacing was commonly used to give a filmic look (25 frames per seconds as opposed to 50 fields).
This meant discarding half the picture information. This gave rise to progressive video whereby the camera would genuinely capture 25 frames per second.

The crux of the matter is .... they didn't change the format the video is stored in (i.e. it is still interlaced - the contents of all the odd lines are stored first and then the contents of all the even lines are stored next). This means progressive video will playback properly on an interlaced screen fed via a standard video signal.

To think about this another way consider the following; video is always stored broken down into odd and even lines. With interlaced video a stream of these odd and even lines would be sampled at the following times:

t, t+f, t+2f, t+3f, t+4f, t+5f, t+6f etc

where t is an arbitrary time and f is the duration of a field (1/50th of a second in PAL)

With progressive video the odd and even lines would be sampled at the following times:
t , t, t+2f, t+2f, t+4f, t+4f etc

If the above is true (and to be fair it is entirely guess work on my part) then I wonder in what way toaster makes use of progressive video?

I believe DVE's are normally created with fielding (ie interlaced video) in mind. I'm sure that the Toasters means of handling retimed video assumes interlaced source (to sample the video at a given time it will make use of the fact that the second field is 1/2 a frame ahead of the first field). The variable speed nature of DVE's means they are probably being sampled in the same way.

That was an incredibly long winded way of saying I don't know the answer. But it would be nice if someone could set me straight on progressive video.

pfrench
06-03-2003, 09:44 AM
You are correct, an interlaced camera (99% of whats on the market) records _FIELDS_ of video at:

t, t+1/60, t+2/60, t+3/60, etc.

or every 60th of a second (50th in PAL). The progressive cameras record an instantaneous _FRAME_, so the _FIELDS_ are of the time.

t, t, t+2/60, t+2/60, t+4/60, t+4/60,, etc.

Note however that _ALL_ television monitors display video in fielded motion, Plug your progressive recording camera directly into your monitor (fielded) and do a quick pan or wave a pencil back and forth in front of the camera. Look closely, you will see what looks like interlace judder.

This happens because your eye is _smoothly_ tracking motion changes that occur every 30th of a second (progressive camera), but the monitor is updating the picture twice as often, so as your eye follows the motion, the monitor draws two pictures in the same location.

Progressive Source Editing:

The VT is internall always performing fielded operations, meaning the DVEs, all of the switcher activity, etc. are performed a field at a time. VTEdit I believe is always fielded, also.

You can, however, in VTEdit drop a de-interlacing filter over your entire project and get that progressive looik.

mblade
06-03-2003, 10:21 AM
That all sounds sensible to me!

So what's the progressive option for when capturing and rendering for?

J.

pfrench
06-03-2003, 01:31 PM
sounds like build-in de-fielders to me...

Progressive render sound good for generating video thats going to be displayed on computer monitors (which are all progressive display.)

I only write software, I rarely get a chance to actually use it... B^)

Ryan
06-03-2003, 05:33 PM
So this is what I gather.

If you want to produce progressive video, you have to treat it as if it was interlaced in the toaster. This means that if you used any transition effects or motion it would be an interlaced effect on a progressive video. So you can only use the toaster for compositing, color correction, and fade type effects.

So when you are done you render out of the toaster in fielded mode. So that it wont blend fields on a progressive video. This would blur the image quality if im correct.

Once you have this video rendered out of the toaster that was saved as fielded even though it was progressive, into an avi(or another format), is the file now interlaced or is it progressive.

So that brings me to the real question.
when the toaster renders out an avi, does it save odd and even lines on differant frames when its interlaced. Or is it in fact one frame with all the lines in order, and when you want to play it back on a TV the player plays the odd lines then even lines then goes on the next frame.

If its still interlaced could you wash it through virtual dub or another program or even aura? or does "render as interlaced" mean that it simply leaves the video as is and puts it in an avi where all the info is saved on one frame.

I imagine this would be great to know when using pixel trackers like icarus.

Thanks for the comments everyone!


Ryan

Admitting Ignorance is Prerequisite for Gaining Knowledge.

mblade
06-03-2003, 08:17 PM
That all sounds good for the most part, but...

Fades are still intrinsically fielded, so they should be avoided if you want true progressive video. (during a fade the second field will be more 'faded' then the first field for any given frame).

If you avoid these and anything else that will cause a distinction between different fields of the same frame then I guess you are working 'progressivly'. I think you are correct in saying that you should still export with fields to avoid the loss of effective vertical resolution caused by defielding. As to whether the final file is progressive or interlaced, I would say progressive, although I think this is now a fine line. I don't believe the interlaced/progressive choice sets a flag within the file (so there is no simple way for any software to determine of its own accord if the file is interlaced or progressive).

When in doubt I like to consider working with film footage in a digital system. Films can be thought of as a genuinely progressive (i.e. 25 frames per second) . When this is telecined for use in a video environment the field/frame issue is resolved exactly as mentioned at the beginning of this topic (assuming PAL - I don't even want to know about that NTSC pull-down crap). Thus if you want to stay progressive (possibly to go back to film) you must follow the work path suggested above.

I've got a hunch that it would be possible to make up a set of progressive DVE's. If they were created progressivly they would behave correctly until scaled in time when Toaster would try and interpret them as fields. So whilst they would be of fixed durations they would allow your progressive work to be a bit nicer (i.e. fades etc).

Anybody want to have a go at making some?

Are we chatting _PROGRESSIVLY_ or would I be better standing in a _FIELD_?

Paul Lara
06-03-2003, 08:28 PM
Originally posted by mblade
Are we chatting _PROGRESSIVELY_ or would I be better standing in a _FIELD_?

You are certainly out standing in your field! :D

pfrench
06-04-2003, 09:49 AM
In VT[3] (and possibly VT[2], can not check) you can render in progressive or interlaced from TEd.

vanguard
06-05-2003, 07:18 AM
Progressive scan is one of those things that always confused me about why you would want to use it.

In an interlaced display system the effective frame rate is actually 60fps. (NTSC)

Why on earth would you want to reduce the frame rate to 30fps.

To make things that are in motion "judder" as they go by?

The SMPTE 24fps film standard was constantly being challenged until computers and video were "threatning" the ivory tower of film.

George Lucas (and many other directors) lobbied for years to get the film frame rate up to 30fps to get smoother motion. Kodak (who would make the most money from this), formed a group of directors and producers to push the idea. Of course the studios were opposed because of the extra expense.

BTW 24fps was picked by SMPTE because it afforded the minimum film use, without "objectional" strobing of action.

NTSC and PAL received their respective frame rates based on the electricity available in their countries.

Consider this: (nearly to the end of the rant now) Given the opportunity in Lightwave to render in field mode, do you?

Why?

Because frame mode (effectively progressive scan because both fields are the same for each frame) well... sucks. It looks bad.

So, IMHO this whole 24p "feels like the look of film" is as much poo.

Its as goofy as a display with 3:2 pulldown mode. Why on earth do you want to make a picture worse.....

At the theater the movie on DVD was shown at 24fps, then spot scanned at 3:2 pulldown already, so 3:2 is moot. The episode of Star Trek which was shot on 35mm film and then spot scanned at 3:2 pulldown is also moot. It must be to get that episode of "Everybody Loves Raymond", to look like it was shot on film......

OK end of rant

At least IMAX runs at 30fps...

OK I'm really done now.

mblade
06-05-2003, 07:34 AM
All true, but there are some applications where progressive is best!

If you are resigned to using frames(which does give a nice filmic look) then surely it is better to use progressive source rather then a de-interlace filter (which 'makes up' half the picture). Combine this with the context the thread was started with - 'producing video for non-interlaced displays i.e. computers - and progressive video looks very tempting indeed.

As for why would anyone want to make video worse than it is... Surely a blur or a colour effect is making something look worse in the sense of moving away from the original as shot by the camera. I don't think anyone would argue that video looks better without them!

Lightwolf
06-17-2003, 02:57 AM
O.k. we are getting way OT here, but what the heck.
As for progessive editing, there's a bunch of reasons for that:

1) your final will be displayed on a monitor (this includes most plasma screens, TFTs, projectors et al, that behave fairly bad with fields), or is intended as a computer video stream.
2) your final will be printed to film.
3) your original material is progressive, and you want to keep the "look".

Since the T3 dows both, great.

As far as film is concerned, a 24fps film projection isn't as bad as it sounds, since the shutter has a frequency of 48fps (the frames are displayed twice) , which at least reduces the flicker and seems to help the eye somehow (hard to say, but I tend to notice strobing less in the cinema, than on a TV set or monitor with 24fps).

Cheers,
Mike

djlithium
06-20-2003, 05:28 PM
Okay. I thought I would step in and and steer this back into reality land.

Shoot interlaced, edit interlaced, render your file out interlaced, THEN turn it to progressive FRAME based video using an application like Virtual Dub.

I usually ditch the 6 frames extra in NTSC and drop it down to 24FPS for doing streaming video and use a field killer filter or deinterlacing option in Virtual dub for 2 reasons.

1. Reducing the frame rate using a decimation filter in your master file for use in your encoding application forces your encoder to work at that frame rate and properly pull out FRAMES instead of trying to blend frames and other gack which most endoing applications from companies like real, MS (WMV/WME) and Apple generally suck at doing. This reduces artifacts and shreading which will cause the compression to look worse then it needs to be in your final streaming video clip.

2. Frame rates like 29.97 are absolutely useless on a computer. The illusion of motion is achieved at just over 12fps, so double it and you get smooth video. This should be what you feed the encoder. You will get cleaner video, smaller file sizes (and thus more stabel streams) and you can work with larger resolutions for final output because you have the extra data headroom to work with.

Do yourself a favor, get virtual dub. It's the handiest tool you will ever use with your Toaster system for making all kinds of neat video.

Also, if you need help in understanding what you can and cannot do with streaming WMV files and what the proper bitrates should be, email me. I have been doing reseach and development in this area for 4 years. I have a nice little guide that I wrote for advanced WMV streaming set up coming out in Keyframe and I can offer a lot of advice on this subject.



DJ Lithium
www.djlithium.com
[email protected]
(also check out www.kyxpyx.net for the koolest little WMV streaming server out there.... KasterBlaster. Streams WMV, WMA, MP3, MPEG1, ASF, Ogg Vorbis on-demand as well as functions as a live mp3 radio station at the same time if you wanted it to be).

cholo
07-08-2003, 06:05 PM
Why would anyone want to shoot progressive video? Well, for one thing because you lose resolution when you shoot interlaced video. For example, try this: do a slow tilt with your camera in interlaced mode. At a certain speed your fields will be duplicate because by the time the next field is recorded the exact same portion of the image that was occupying the last field now occupies the next one so that you end up with stairstepped borders because of this. Once you've noticed this in one shot you'll start noticing it in more shots as you view them. You may gain in temporal resolution but you end up with less resolution as a result. If video were 60 frames instead of fields per second that would be a whole different story. The next thing is that since you have much more temporal resolution camera moves tend to be much more sensitive to shaky motion and less forgiving than lower frame rates are. Last but not least in progressive you end up with a product that is compatible both with video, computer displays, etc... and not only are you more compatible and flexible, but you also end up with smaller file sizes and better quality in the computer since you can achieve better looking images at 24 fps rather than 30 fps at any given bitrate. In other words, less frames equals more bits per frame to preserve quality. Aside from all this is the whole aesthetics of the strobe like motion of 24fps, etc... but that is a more subjective approach to the whole subject. I'm about to do a 24 fps project on the toaster and the approach I had thought of was the following... Capture all my footage interlaced, remove pulldown and convert to 30fps clips for editing (it'll run faster so audio must be processed in the same way to remain constant), if I use any dissolves in the edit I'll apply a deinterlace filter over the transition segments, as for PIP effects and such I find it better to do them in After Effects so as to be able to apply motion blur to them. Next I'll render the whole program as a single clip and in AE convert it into a 24fps AVI file with the original audio track back. Last, compress into mpeg2, burn into DVD, playback, enjoy! :)

Hope any of this helps.

Keith Nealy
07-22-2003, 12:47 PM
One of the biggest advantages in using progressive as I understand it is in keying. As so many major motion pictures use keying for just about everything, you can get much smoother, less troublesome keys of a frame than you can on a field.

Just look at the amount of keying on Phantom Menace.

hamilton
08-17-2004, 10:22 AM
I want to get an XL2.
Will VT3 allow me to use the different scan rates the XL2 now offers?
There seem to be no options for importing 24p, 30p etc. in the capture menu.

Michael Hamilton

chuckd
08-23-2004, 07:39 PM
I have just done a couple of projects in progressive and, frankly, I like it. Yes, there are things you need to be careful about doing when panning but if you follow some basic guidelines (that are printed in the ASC Cinematographer's Manual) then you should be ok.

I learned a few things with progressive:

1. DON'T use an XL-1 in progressive mode!

2. DON'T use an XL-1 in progressive mode! Ok, ok I am repeating myself. Seriously though, the way the camera does progressive is a rip off. It takes one field in your frame, throws it away and then copies the other field onto it. You lose half your vertical resolution!

For me that's completely unacceptable. It makes any shot that's not in bright sunlight look out of focus and waaaaaaay too soft. If you compare the XL-1 with the Panasonic DVX-100 side by side you'll end up sending the XL-1 back to Canon in little pieces. (Frankly I'm not impressed with the XL-2 either).

3. If you don't want to shoot progressive but want to work with it in VT there is a de-fielding filter that does an excellent job in the Effects folder. This is my preferred way of doing progressive when shooting with a normal camera.


But it would be very nice if VT would support 24p frame rates. I don't see why this would be difficult-- it can convert between PAL and NTSC on the fly. It can't be any more difficult to do 3:2 pulldown in realtime.

I would love to see support for Panasonic's 24p-A format!

Tom Wood
08-24-2004, 08:35 AM
Mind if I ask a question here?

I am doing character animation like my avatar over there with the end result being a DVD. I've found that rendering out of LightWave in fields-off dithered motion blur gives me the best looking images in a tilt motion, but in a pan motion using fields works best. (For a NTSC monitor) I'm tempted to suggest that the DVD should be displayed on a progressive scan device and just forget about NTSC field problems.

My workflow is:

LightWave>Mirage>VT3>TMPGEnc>DVD Workshop 1.3

Since I use the third party TMPGEnc plug-in, I never render anything out of VT3, I just drop the VT-Edit project into it. Even though I'm rendering out of LightWave in non-fielded, and compositing in Mirage in progressive scan, would I be better off adding the de-fielding filter in VT3 as well to ensure that it is passing everything on to TMPGEnc in progressive scan mode? Or would that actually reverse the effect?

Thanks,

TW

Pete Draves
08-24-2004, 08:44 AM
the panasonic has a 30p mode that looks great.
I demoed that camera 2 years ago at nab for panasonic.
the new model has anamorphic 16-9
great camera
Pete

bradl
08-24-2004, 06:36 PM
I am getting proficient at making DVE's (at least the monotonous compiling part) as I have just finished making about 200 of them. Soooo... I could look into what it would take to make a progressive DVE effect, fade for example, and attach my attempts here for evaluation.

I may need a little discussion on the finer points after I look into first. For example, normally, the LW DVE frame rate (FPS) needs to be set at double the field rate (120 FPS for NTSC and 100 FPS for PAL). Apparently this is because it takes two complete fields to make one DVE 'field'.

I think it breaks down something like this: the first field holding the Warp data and the second containing the Alpha. One unique twist to the setup is you actually turn on Steroscopic Camera rendering. Each field rendered in LW outputs two files. For example the Timeline for NTSC is set to 240 (PAL 200) frames but the final folder of files is actually 480 (PAL 400).

After compiling the DVE you end up with a DVE that is 120 frames long (4 seconds). Soooo... the 2 files (via Stereoscopic) combine to make one field and two of those make one frame. 480 down to 240 down to final of 120. Clear as mud?

Any comments or suggestions as to how all this may relate to Progressive, if it is even possible?

bradl
08-24-2004, 07:17 PM
OK, I have already thought of a way to do this. May have to try two versions to line up field dominance of the DVE to the Progressive RTV. If off, you would get one field of the progressive frame fading in one 'progressive' DVE frame and the complementary field jumping to the next DVE frame to fade and that would look very bad (in theory).

chuckd
08-24-2004, 08:56 PM
Tom, I don't see any need to use the de-field effect in VT Edit. Since you aren't bringing any fielded info in it doesn't matter that VT will send the images out fielded-- the two fields it creates are not temporally displaced. So you're getting no benefit from it.

But you have noticed that panning takes a bit of finess to get right in progressive footage. Unfortunately that's the big trade off with progressive formats.

Hve you considered rendering out your animations at 24p and then doing 3:2 pulldown in Mirage? You might get the look you're after without some of the drawbacks of 30p.

pfrench
08-24-2004, 09:13 PM
apology in advance for the un-edited rambling answer


Originally posted by bradl
I am getting proficient at making DVE's


Awesome!!!



I may need a little discussion on the finer points after I look into first. For example, normally, the LW DVE frame rate (FPS) needs to be set at double the field rate (120 FPS for NTSC and 100 FPS for PAL). Apparently this is because it takes two complete fields to make one DVE 'field'.


The DVE Frame rate needs to be double the field rate so the DVE can be "played" inside the switcher and have smooth, fielded motion at 1x and 2x field rate (slow and medium TBar speeds)

Also, the DVE is [i]FRAME RENDERED (NOT field rendered)[i] so the switcher can "park" the TBar during a transition and you don't get fielded motion. It looks solid.





I think it breaks down something like this: the first field holding the Warp data and the second containing the Alpha. One unique twist to the setup is you actually turn on Steroscopic Camera rendering. Each field rendered in LW outputs two files. For example the Timeline for NTSC is set to 240 (PAL 200) frames but the final folder of files is actually 480 (PAL 400).


Uhm.... no. B^)

stereoscopic rendering is a sneaky way of doing "two pass" rendering inside LW so the DVE Compiler can do a correct, photorealistic analysis.



After compiling the DVE you end up with a DVE that is 120 frames long (4 seconds). Soooo... the 2 files (via Stereoscopic) combine to make one field and two of those make one frame. 480 down to 240 down to final of 120. Clear as mud?

Any comments or suggestions as to how all this may relate to Progressive, if it is even possible?

Ok, DVEs are supposed to be eight seconds long (240 frames NTSC, 200 frames PAL) so they can "play" as fielded within the switcher.

bradl
08-25-2004, 06:39 AM
Originally posted by pfrench
Ok, DVEs are supposed to be eight seconds long (240 frames NTSC, 200 frames PAL) so they can "play" as fielded within the switcher. OK... but I have no 8 'second' DVEs in my system? Are you saying a DVE 'frame' consists of 4 fields at normal playback speeds?

bradl
08-25-2004, 06:50 AM
Are you also saying each frame by nature a Progressive frame (at 100% speed)? If so, at speeds faster than 100% is this 'Progressive' framing defeated in reality?

pfrench
08-25-2004, 11:00 AM
Each "Progressive" Frame in the DVE contains a DVE for both the odd and even field at the same instant in time.

When a DVE is performed in the switcher, the slowest it should be played is at four seconds (120 frames NTSC, 100 frames PAL), assuming the DVE is 240 (200 PAL) frames long. This way the switcher uses a different instant in time for each field, giving smooth, fielded motion.