Progressive scan questions

John_Cline wrote on 5/24/2004, 10:19 PM

This is probably going to turn into a long post...

There are two types of resolution; "spatial resolution" which, in the digital world, refers to the number of vertical and horizontal pixels of image information. In the case of NTSC DV it's 720x480. The other type of resolution is "temporal resolution" which is how many frames are captured in a given amount of time. In NTSC DV, it is 29.97 frames per second. However, DV is interlaced which means that individual images are actually captured 59.94 times per second. These are called "fields" a single frame contains two interlaced fields. The first field is captured at the full 720 pixel horizontal resolution, but only every other vertical line is captured. During the second field, which occurs 1/59.94th of a second later in time, it once again captures the full 720 horizontal pixels and the other half of the vertical lines that it didn't capture in the first field. Essentially, NTSC DV has a spatial resolution of 720x240 and a temporal resolution of 59.94 images per second. While it has relatively limited vertical spatial resolution, the high field rate captures motion very smoothly. Progressive video has 1/2 the temporal resolution of interlaced video, since there are only 30 discrete images captured per second. The temporal resolution of 24 fps progressive is even worse at 24 images per second. There is an effect called "persistence of vision" which is where the brain perceives individual frames of either film or video as continuous, smooth motion if the frame rate is high enough. There is some debate about how high is high enough, but most people agree that it is a mininum of around 40 images per second. 24 frames per second is NOT high enough, look at a movie where there is a wide, panorama shot which pans fairly quickly, it seems to stutter because we can perceive the 24 frames per second as indiviual frames and not smooth motion. 30 frame progressive isn't quite fast enough either, there will be certain shots where it will seem to stutter.

There are a couple of problems I see with taking interlaced video and converting it to progressive video. First off, you blow of half of your temporal resolution. Secondly, without a more sophisticated deinterlacing algorithm than the one contained in Vegas, you don't gain any more spatial resolution than the 720x240 you started with. You can interpolate the frames, which throws out one field altogether and then guesses what to fill the "in between" line with depending on the image information contained in the line above and below the one you are trying to interpolate. You can't really create resolution that wasn't there to begin with. The other method of deinterlacing in Vegas is "blending fields" which has the effect of blurring the image in time. In my opinion, neither of these are acceptable. (The Magic Bullet Suite software actually does a pretty decent job of converting 29.97i to 29.97p or 24p.) OK, so now you have progressive video, but it has no more spatial resolution and half the temporal resolution. Now if you are shooting high definition (HD) video at 1280x720p, you have a spatial resolution of 1280x720 pixels at a temporal resolution of 60 progressive frames per second. This looks pretty damned good, both from a spatial and temporal resolution standpoint. 1920x1080i looks pretty stunning as well.

When photography was first invented, people complained because it didn't look like a painting, which is what they were used to. So, for quite some time, photographers tried to make their photographs look like paintings. Now people are complaining because video doesn't look like film. Yes, 24p and 30p has a certain "once removed" feel because of the limited temporal resolution and that's fine for a dramatic effect. But that's just because that's what we're used to. Really, the trick is to light your video and compose your shots like you are shooting film. Personally, I really enjoy watching a video which has been lit like film and has high spatial resolution AND high temporal resolution. One thing that most video really lacks is that you can't do dramatic depth of field tricks because of the physics of camcorder lenses and the small size of the CCD imaging chips. You just can't get really shallow depth of field with video, even with the lens aperture wide open. Film has it all over video in this regard and there is no software that will fake this for you.

The semi-short answers to your specific questions are:

Progressive video which is converted from interlaced video using the tools in Vegas has no more spatial resolution than what you started with and less than half the temporal resolution. This doesn't sound "better" to me.

Should you be "adjusting settings so that all projects are made in progressive scan?" Yeah, but only if you want to throw away over half the information for dramatic effect.

even some of the older DVD players that don't support progressive scan can still play your video. Is this true?

Yes, a lot of Hollywood features released on DVD are 24p, but the DVD player does "3:2 pulldown" which converts the 24p video to 29.97i. This 3:2 pulldown has it's own quirk in NTSC land, 3:2 pulldown displays a film frame for 3 video fields, then the next film frame is displayed for 2 video fields, then the next film frame is shown for 3 video fields, etc. This alternating between 2 fields and 3 fields produces the necessary 2.5 multiple of 24fps, thus producing 60 fields per second. However, some video frames will have have two identical fields and some frames will have two fields which are parts of two different film frames. So, 3:2 pulldown has its own strange temporal stutter to it. Anyone that has ever watched a movie on TV has experienced this and "that's what were used to." 24p can only be properly displayed on a TV capable of 24p using 24p source material and a device which outputs 24p, like a progressive capable DVD player. If you make a DVD at 24p and view it on a regular TV, the DVD player has to insert the 3:2 pulldown and you're back to the odd stuttering field rate I just described. In PAL land, they just speed up the film from 24fps to 25fps and the audio and video plays a little faster than it should. Once again, incorrect though it may be, that's what the PAL folks are used to.

Are there any advantages at all of staying with interlaced video?

Unless you are trying for some effect, then yes, staying with interlaced video will maintain all the spatial and temporal resolution that you captured in the first place.

Personally, I don't want my audio CDs to have the tick, pops, rumble and inner groove distortion of a phonograph record, nor do I want my new Acura 3.2TL to drive like a 1962 Volkswagen Beetle. And I sure as heck don't hear anyone saying how much better analog video recording looks than digital video. (Curiously, unlike the audio business, there is no such thing as "classic" video gear. I've got a Sony BVH-2800 1" Type-C video recorder sitting in my edit suite, even though it originally cost over $40,000, it never gets used for production because it sure doesn't look "better" than my considerably less expensive digital video gear.) You know, I'm all for film style lighting, shot composition.and storytelling, but perhaps it's time we stopped trying to make video look like 100+ year-old film technology (at least as far as temporal resolution is concerned.)

You all may flame at will...

John

wolfbass wrote on 5/24/2004, 10:32 PM

John:

Thank you for that in-depth and informative reply.

Andy

Phil_A wrote on 5/24/2004, 11:29 PM

A brief corollary to John's 'soapbox' (and it was impressively articulated):

Check out the DVD movie "Once Upon A Time In Mexico." It was not until after viewing that I discovered it was shot on video (HD) and not on film. Through the entire motion picture viewing, I didn't stop once to consider that this was not film... And it looks stunning!

Videot wrote on 5/25/2004, 6:28 AM

Where are a great many films around that were shot entirely on video. the most technically challanging would have to have been 'Russian Ark'. The 90 minute film was shot in one take & involved thousands of extras & covered several kilometers.

farss wrote on 5/25/2004, 8:06 AM

Russian Ark was not your average video shoot though, the cameras come close to 70mm film in res. Apart from that though many films today go through a DI which is I suppose technically DV. Not something that goes to tape and scanning is probably done at seconds per frame but it's still a digital process. What really bugs me is the latest trends in film. I have no idea what silver retension, bleaching or removing the anti halation layer does but I know the result looks more like it was shot on a clapped out first generation VHS camera.

TVCmike wrote on 5/25/2004, 8:22 AM

John, I think there's one more issue that you somewhat glossed over, and that's the suspension of reality that 23.976/24fps can give versus 29.97fps. For some productions, this is important in terms of its look. Granted, you should try to stay within integer multiples of your original source if you want the best results. In general, the faster the framerate, the more "real" something will look. This is not necessarily what you always want, especially in a creative production.

As a side note, 70fps is approximately the limit of human vision temporally, and (IIRC) 4000 horizontal lines of resolution is our spatial limit. I know the temporal limit is pretty certain because there has been a discussion of it in the gaming world as far as what frame rates a video card should support. It'll be a long long time before we can edit video in those resolutions.

To Videot...

Yes, some older DVD players don't fully support the progressive modes but still hack out something. Only the baseline resolutions of the original spec as outlined in the DVD Demystified FAQ are fully supported in all players. If high compatibility is your aim, stick with the "standard" and good compatible media. Also, don't forget that most Hollywood movies are shot on progressive digital or celluloid formats, whereas the vast majority of independent video producers are using DV cameras.

And on that note, if you can help staying in your original field rate and resolution, you'll have the best preservation of the original video. I hesitate to say "quality" because the qualities of the video are purely subjective manifestations IMO. I encourage you to try using various field/framerates with your own videos and see. In fact, you might be surprised what a lower frame rate look might do for your clients' perception of your productions. We can all tell you what we think, but you'll only know by trying. So Just Do It™.

John_Cline wrote on 5/25/2004, 9:56 AM

TVCmike,

Well, I did sort of gloss over the "suspension of reality" aspect. I called it the "once removed feel", but that's essentially what I was talking about. And, yes, I agree wholeheartedly that it is extremely important in terms of "look" for certain productions. I referred to it as "dramatic effect" several times in my post.

Thanks for adding your informed perspective to this thread.

John

Bill Ravens wrote on 5/26/2004, 1:10 PM

Kudo's John. Your essay puts very well what I've been saying about film vs DV for 5 years. Glad to read the thoughts of a reasonable man. I will add that the use of 35mm lenses on a DV camera, like an XL1s, goes a LOONG way in resolving the DOF situation you described above. Unfortunately, there is a major hit in image magnification with the adapters.

John_Cline wrote on 5/27/2004, 9:17 AM

Thanks for your kind words, Bill.

You know, I really expected to take more heat from the 24p camp here on the forums. I guess it was probably obvious from my post that I just don't quite understand why there has been this sudden fascination with 24p. Yes, there are certain productions that would benefit from the "suspension of reality" effect, but it seems like "regular" folks somehow think that all of their productions need to be 24p to make them magically look more "professional."

Just for the record, when it's appropriate, I will use Magic Bullet in After Effects to de-interlace some of my stuff and make it 30p and, perhaps, even use one of the Magic Bullet "Looks" to enhance the effect. Of course, all of this works much better if the video has been lit using "film style" lighting techniques. The de-interlacer in Magic Bullet is "adaptive" so that in parts of the image where there is no motion, it will maintain the full 480 lines of vertical resolution. There are also a number of filters for VirtualDub, like Donald Graft's "Smart De-interlacer," that will do this as well. The de-interlacing algorithms in Vegas don't do this.

There was an article written in one of the production magazines a while back that explained very well why DOF is so difficult with small chip video equipment. I'm almost certain it was also posted on the web, so I'll see if I can find the link and post it here.

John

Edin1 wrote on 11/20/2005, 3:37 AM

I have to say that I like your detailed explanation, John, but there is one thing (actually two) that I have an issue with:

1. I agree that interlaced video loses half of its vertical resolution when there is a lot of movement, but whenever there is no movement, the interlaced video has the full spatial resolution.
Now, to beat my own argument here, the point of the video is to have motion, otherwise it's just a picture ;-)

2. Since I heard the argument about the 24p being associated with dreamlike (unreal) look, and 60i looking more like reality, I had trouble accepting it, and kept thinking how, what, etc.
Then after some time watching a few films and videos, and comparing them with how our (my) eyes see, I have to say that I don't agree with the dream/reality explanation, and have to say that I have a better one, mainly focusing on motion and blur issues.
Here it is:
VIDEO, INTERLACED (50i or 60i)
We perceive video as something distant, viewed from the outside, and not being in it. The objects in video can be seen better while it is being panned, the movements are more liquid, reflecting realistic look, and not having the stutter which immediately betrays it as video/film. And when there is movement, the image is more clear, and we can focus our sight onto almost any object in the video.
Objects in video look like in real life, we can focus on them when and where we want.

FILM, PROGRESSIVE (24p)
Seeing things on film is like seeing them through our own eyes (yes, I'm talking about being awake).
I have noticed how things get blured while I "pan" my look around, how things sometimes look stuttery when I move my gaze (eyes) faster, and it looks pretty much like film. And when I am following someone/something moving fairly fast, the background gets blurred.
Also, depth of field seems to be very much like in the movies, where what we are focused on remains clear, and everything else dull, blurred, out of focus, even if there is very little distance between them. OK, the depth of field trick can be done with interlaced as well, so this is not the one affecting the film/video difference much.
The main difference between film and video (progressive and interlaced) is the way they treat motion.
The video looks like the reflection of reality, our surroundings, environment (because of smooth motion), just something to be gazed upon, while the film looks as though we are seeing it with our own eyes, which has that dramatization effect, and making us feel more involved personally into what is happening on the screen.

farss wrote on 11/20/2005, 4:25 AM

Sorry but there's some really dodgy logic in there!
If as you suggest we naturally blur fast moving objects (and I don't even agree with that!) then if a projected image moves quickly will our eyes not naturally add the appropriate matching amount of blur?

Given that and the variability of humans wouldn't it be more natural to have no MB in the image to start with and let the viewers eyes / brain add there own?

As to DOF, well yes, when I focus on something very close, distant objects move out of focus however I can shift my focus, if the projected image has all but the subject out of focus there's no way I can focus on the villan in the background. Hardly an imitation of a nature.

Human visual perception is a very complex beast, the amount of data sent from the eyes to the brain is remarkably small compared to the amount our eyes receive however we do gradually build an incredibly detailed image, something equivalent to 100s of megapixels over a huge field of view, I think it'll be sometime before we come close to presenting an image at anything like the limits of human perception. However understanding how our visual perception works gives a clue as to what to avoid when presenting moving images. Continual full screen motion can be seen to be very bad, out eyes and brains are forever trying to fill in details but it can take minutes for the full picture to develop in the brain, moving it around frustrates what the brain is trying to achieve leading to stress.

The whole problem with the 24p vs 60p argument is conditioning, there's some remarkable examples of how this affects how we react, even our emotional response to certain colours varies across cultures. Mostly 24fps is shown on large screens at very high resolution, most people see 60i at low resolution on small screens. To truly evaluate the difference you'd need to say shoot and project 35mm film at 60fps, ideally using the same scenes.
Of course as far as I know none of the modern display systems actually show 60i, so the real issue is I think going to become 24p Vs 60p which is a somewhat different argument.
Oh and by the way, you can shoot 24p with less motion blur, it's done a bit to produce a feeling of tension due to the more stroby nature of the motion.
Bob.