Is IVTC appropriate for a VHS source?

[mcpogue]

Is VHS material tc'd or is it 30fps all the way through? In other words is ivtc appropriate?

[edDV]

Film source on NTSC VHS would have 2:3 sequence just like TV.

It's up to you whether IVTC is worth the bother. If the target is a TV, a HDTV hardware deinterlacer will probably do a better job.

The only benefit to doing it before DVD authoring is to squeeze in 20% more minutes at a given quality.

[FulciLives]

NTSC VHS is always going to be 29.970fps

Some material will have a "standard" 3:2 pattern. Such material can be Inversed Telecined back to 23.976fps

Some material will be interlaced through out (usually stuff shot on video) or will have such a corrupt 3:2 pattern than doing an IVTC will do more harm than good.

I'd suggest installing AviSynth and playing around with the decomb plug-in which comes with some very good reading material.

- John "FulciLives" Coleman

[adam]

Edit: You guys type too fast.

It totally depends, some is some isn't. The only way to know for sure is to view it frame by frame and see if its pure interlaced (nothing but fields visible) or whether there is a set pattern of progressive frames to fields, in which case it may be able to be IVTC'ed. For this I use TMPGenc. Just load it and use the preview option or one of the filters such as deinterlace. Scroll through it frame by frame. What you are hoping for is 3 progressive frames followed by 2 fields. Check it throughout to make sure it follows the same pattern always. If so, its a perfect candidate for an IVTC and just use any automatic IVTC filter. If it uses a different pattern then it is still possible to do an IVTC but maybe not worth the trouble.

My experience is that most commercial NTSC VHS sources IVTC just fine.

[adam]

edDV, I think the 20% less frames number is kinda meaningless personally since the frames being omitted are highly redundant. They are after all derived from fields/frames that you do still keep.

But in my experience there is a tremendous quality gain to be had by encoding using frame pictures versus field pictures. Regardless of what encoder I use, I have to increase my bitrate substantially when doing interlaced footage versus progressive or the quality just isn't there. So I really think that there is always a benfit to IVTC'ing where applicable, not just because of more bits per pixel but even more so just because of the nature of encoding interlaced footage.

[mcpogue]

Could someone plz post some pics so I know exactly what I'm looking for? i.e. progressive frames vs fields...thx

[adam]

A progressive frame will just look like what you'd think a frame would look like. Its just a whole picture, a snapshot of your footage.

A field will be if you took that frame and selected every other scan line. So it looks like a comb. Like this.

[edDV]

Read this and all will be clear.

http://www.dvdfile.com/news/special_report/production_a_z/3_2_pulldown.htm


Top row is the original 24 fps film, run at 23.976 fps

Second row is "telecined" NTSC video as broadcast and recorded on VHS, or 8mm, or laserdisc or DV, or DVD (interlaced). An extra field is inserted after every four field sequence to bring the field rate up to 59.94 fields per second or 29.97 frames per second.

Third row shows IVTC (inverse telecine) as done in software or by a hardware "cinema" deinterlacer in a progressive DVD player, progressive HDTV or progressive TV projector. Result is a recreation of the original 23.976 fps sequence. This is what gets stored on a home made "progressive DVD"

Last row shows how a progressive DVD player, HDTV or TV projector repeat these frames in a 3:2 pattern to generate a 59.94 frame per second progressive display. Yes 59.94 frames per second.



HDTV film broadcast uses a similar telecine process for 1080i/29.97 interlaced. The telecine process, IVTC and frame repeat for progressive display is the same as for NTSC and 480p except for frame size.

Progressive 480p/576p and 720p can be directly broadcasted at the 23.976 frame per second (25 fps in PAL areas) film transfer rate. In theory 1080p/24 is can also be broadcast but this not currently being done. The HDTV will then frame repeat for 59.94 frames per second display.

1080p/23.976 or 1080p/59.94 or 1080i/29.97 will be the standards used for HD DVD (also 720p/59.94 or 29.97 or 23.976). Film sourced material will use 23.976 same as current DVD.

[lordsmurf]

Most of the time, no, it's not appropriate to IVTC video.

Your only real exception is "big name movies" that are completely uncut and released commercially. Something like Terminator, Rambo, etc.

Most things you get off tv broadcasts, especially satellite, have been edited in the interlaced domain.

Satellite suffers an additional "flaw" in that interlacing patterns can change midstream without notice, usually at a commercial break. That's just the nature of the MPEG TS system.

Many tv shows may be filmed, but the final editing and effects are done interlaced (Stargate SG1, for example).

And since "big name movies" are generally the only ones, you could probably just as easily buy the DVD version (if it exists, of course).

A forced IVTC tends to leave "funny looking" artifacts on video. Chroma ghosts are very typical, where their is an after-print of red, even though the luma portion of the signal seems to have IVTC'd okay.

I've forced a few times for menu work, and it's "good enough" but the menu is only 5 minutes or so long at most. I don't know that I'd want to watch damaged video for the whole length of a movie. And the larger the tv, the easier it is to see the flaws.

[edDV]

I agree that software IVTC needs to be hand crafted if used. Any errors that get by you are encoded into the video.

Hardware IVTC ("cinema deinterlacer") does a realtime analysis during playback. It reacts to on the fly to edits and mixes of film and camera interlace.

Current hardware deinterlacers aren't perfect but any errors generated only affect the current playback. The video file is not touched. Future hardware will improve future playback.

[mcpogue]

My reason for asking in the first place is I want to put movies longer than 90 min on a 4.37 GB disc which naturally means a reduction in bitrate, i.e. <6Kb/s and I'm looking for ways to reduce the overall amount of data so that the quality of the output will be "decent" at an anemic bitrate. The only way I know to accomplish this is to: 1) reduce the frame rate. 2) reduce the frame resolution.

[lordsmurf]

Two hours will be fine quality unless it's super high action close-ups of some sort.

720x480 @ 5500k VBR with 8000k max + 256k AC3
352x480 @ 4500k VBR with 5500k max + 256k AC3

[FulciLives]

Two hours will be fine quality unless it's super high action close-ups of some sort.

720x480 @ 5500k VBR with 8000k max + 256k AC3
352x480 @ 4500k VBR with 5500k max + 256k AC3

I do so love the Lord of Smurfs but this isn't the first time I've noticed him giving strange bitrate calculations.

Here is the bitrate calculator from this very site:

https://www.videohelp.com/calc


As you can see you need to lower the bitrate from the suggested 5500kbps to 4820kbps and even then if you want room for DVD authoring overhead I would go as low as 4600kbps (to be "safe").

I don't understand how he thinks an AVERAGE of 5500kbps with 256kbps would fit 2 hours on a single DVD recordable.

I understand what he did with 352x480 ... he is saying that 4500kbps is fine and you don't need to go higher and this is mostly very true ... in fact for 2 hours I would probably just do a CBR of 4600kbps if using Half D1 resolution (352x480). The MAX bitrate for Half D1 is around 5000kbps but for the most part it is hard to tell a difference between 4500kbps - 5000kbps with Half D1 resolution.

- John "FulciLives" Coleman

[lordsmurf]

720x480 @ 5500k VBR with 8000k max + 256k AC3

This is what a DVD recorder does in SP 2-hour mode. I've also done it many times in years past in software.

A couple DVD recorders do 5000k average, but they always stop at 4GB on the disc. The other spec runs out to about 4.3GB.

There are a number of set times that I don't need a bitrate calculator, I can do it in my head based on experience. Plus bitrate calculators fail to account for VBR, drops and rises. I don't have much faith in those calculators. Mathematically sound or not, they screw up a lot.

If you want to feel safer, go ahead and drop the max from 8000 to 7000, but I'd NEVER go less than 5500k on 720x480, I'd just assume drop the res before slipping into macroblock land.

[gammer]

As I am an avid ATI AIW user like Lordsmurf, I would like to mention that capturing with ATI's MMC tends to endcode a bit lower bitrate then what you specify. So setting 5500k may very well get you 2 hours and fit on a 4.3GB disk with good quality.
Another thing I love about MMC is that if you are capturing somthing that is widescreen, then MMC doesn't waist bitrate on the black bars; it only uses bitrate for the picture. The first time I captured a widescreen movie I set up my bitrate and was expecting it to be close to 4.3GB when the capture was done, but when I looked at the file size when completed it was only close to 3GB in size.
I know that if you specify a bitrate in TMPGEnc on a movie that has the widescreen bars, it will use the bitrate for the hole screen as apposed to just the picture like MMC does.

[adam]

Hardware recorders are not going to perfectly match the bitrate settings you use, whereas a software encoder will. Use those numbers in anything else Lordsmurf, and your file will be oversized. There is no way around it, 2hrs at 5500kbps will not fit on a DVD-5 even with no audio at all. Doesn't matter what you use to encode with.

Since we don't know if he is using the same DVD recorder as you, or whether he is even using one at all, it would be best for him to use a bitrate calculator. I don't see what VBR has to do with it either. Bitrate can fluctuate as much as you want, all that matters is that you specify the correct average. The correct bitrate to use is an absolute number that can be calculated down to the sector. Whether or not your encoder maintains the settings you use is another question, but I find software encoders to be terribly accurate. If you are getting 2hrs at a setting of 5500kbps on your DVD recorders then obviously it is not accurate.

[lordsmurf]

A Matrox DV stream encoded to MPEG, out from a FCP timeline using Quicktime Pro, will give 2 hours with a setting of 5200k VBR. I've seen this done at least 45 times now, like clockwork. Fills the disc perfectly, and even has a nice motion menu using about 50MB.

Bitrate calculators simply do CBR, it's not valid for VBR encodes.

Something like Procoder also does well. It's your wasteful encoders like TMPGEnc that have problems keeping size down.

[adam]

Have you ever checked the bitrate of the encoded file afterwards to ensure that it was in fact 5200kbits? If it fit as you say than I can assure you that it did not have an avg bitrate of 5200kbits regardless of what settings you used. There is no arguing with math.

As for bitrate calculators not working with VBR all I can say is that you must be making some sort of mistake. Even CBR does not have a constant bitrate.

mcpogue just do yourself a favor and calculate the bitrate you need yourself. If you need a bitrate calculator use the one on this site or use FitCD.