Best choices for 720x480 Interlaced source?
 

I am converting files that originated from ReplayTV. The source is always 720x480 Interlaced (with a HUGE bitrate). Even if the content is movie, captured VHS, TV Show, Hi8 movie... it still produces the standard 720x480 high bitrate interlaced source.

What makes the most sense for making high quality MPEG1 files that will mainly be played back on TV (not HDTV) via S-Video input.

My thought was the 352x240 (or 480) did not create as sharp a picture as the current KVCD3 template but that 576 was too many extra lines of resolution and it was just wasted space.

Also, all the original files are fullscreen 4:3 TV. I would prefer to keep them fullscreen (without black bars) but I don't want to encode more data than is necessary so the file size is smaller.

Final issue is whether it makes sense to deinterlace or not. Since the playback is interlaced is there value in one way or the other?

I have done several tests but I prefer to understand the science or logic behind settings. You can not always trust your eyes on a test since the content can also effect the perception of quality.

Any help or theories on this subject are welcome.

Grantman

Hi Grantman,

I would go for 352x480 resolution. You also might want to find out if the capture you are to convert was telecined. If it was, then you should IVTC with Telecide() and Decimate() in a .avs script. It you are encoding to MPEG-1, that's all you have to do. Unless the material was not telecined, then you must deinterlace and encode at 29.97fps. If your target is MPEG-2, and the material was telecined, then you should also IVCT, but you encode with 3:2 pulldown. If it was not telecined, then you encode MPEG-2 interlaced at 29.97fps. Just keep in mind that anything you encode at 29.97fps will not look as good as a 23.976 encode :!:
But then, you can only encode at 23.976 IF the material was originally shot FILM (24fps). Hope this helps 8)

-kwag

Why 352?
 

First a couple of clarifications...

I am making MPEG1 files only. The content is not a movie. It is either VHS capture via ReplayTV inputs or actual cable TV recorded on ReplayTV.

I am making the assumption that non of this is telecined. Is this a bad assumption?

I am not going to 23.97 because it always looks a bit off since you are throwing out those extra frames. (I don't mind the file size since this is mainly 1 hr or 30 minute items).

Why go with 352x480. I am not doubting you just trying to learn why.

I did some tests tonight with 352x240, 352x480, 480x480, 528x480.

As expected it seemed that each higher resolution looked better. I kinda understand that 480x480 would look better than 352x480 since 352 is not quite enough to saturate an S-Video input. However, why isn't 528 just overkill on 480?

Maybe I am focusing on the wrong point but my understand is that at the end of the day you are going to be limited by the resolution but that you are wasting space to include extra resolution?

Is there an easy explanation?

Grantman

Telecined?
 

Is there an easy way to tell if something has been telecined?

If I caputure a VHS tape that is a concert or music video interlaced at 720x480...

Can I tell from that file if the contents of the original VHS tape were telecined?

Grantman

One way is to Look at the frames one by one in Vdub. You should see 3 solid (progressive) frames and 2 interlaced in a repeating cycle.
ozjeff99

Re: Why 352?
 

Yes it is :!:, because it depends on the original show that was recorded. Some may be telecined and others not.No :!:, if the original is 23.976, and it has been telecined for transmission, you'll get better qualty by IVTC'ing and encoding at the original frame rate :!: Because even though your recorder captures at 720x480, that was not the original frame size. Specially from VHS. So 352x480 would be a good average resolution for encoding. You'll probably max out your quality at around 480x480, but not any higher.For the same reason that there's no use on encoding at a higher resolution than the original material. Exactly :!:, as described above. Just did 8)

-kwag

Just confirming Kwag was correct... (it was never in doubt)
 

I did more testing and it turned out exactly as you suggested. Thanks for the telecine reminder. I have tried several pieces of content that I would not have thought would be telicined and it looks like they were. (I thought only movies were originally 23.97).

The #1 rule of digital video or audio is that you can not create information that was not there to start. It is tempting to try and make use of a 720x480 source but if it is a capture of VHS then it is silly to go above 352x480 since the content was around 352. Test concluded that to be true.

Thanks Kwag.

Grantman

How to determine telecine?
 

ozjeff99, thanks for the idea about looking in VirtualDub at frames.

I did that with one of my VHS captures and I found the following pattern:

BBPBBPBBPBBPBBI

I am not sure what that would be? I am guessing it is NOT telecined? Any comments?

Grantman

Re: How to determine telecine?
 

That's the IBP GOP structure. That's not what you're looking for :!:Ignore the IBP, and slowly move frame by frame. You should see three consecutive frames that show no overlap/interlace, followed by two frames that look interlaced. Look at this site so you know what I mean by "interlace" look :)
www.100fps.com

-kwag

Thanks... I got it now.
 

I was tired and trying to make it more complex. If you look at the actual pictures it is obvious when the item is telecined.

Thanks,

Grantman

Hi Grantman

I am also experimenting with "high quality" ReplayTv files. I have not been successful in IVTC'ing any of them. I've just resorted to encoding at 29.97 interlaced to Kwag's full DVD template. I've only done a handfull of Battlestar Galactica episodes but so far all have turned out great. Only minimal quality loss when compared to the original.

Best of luck!

Those are some of my favorite too ;)

-kwag

Is this Interlaced?
 

I just walked frame by frame through a nice clip from The Simpsons. Homer is walking across the screen so his white shirt is perfect for showing interlace lines.

Using P=progressive (no interlace lines at all) and I=interlace (very visable black lines) I have found the following pattern.

IPPPPIPPPPIPPPPIPPPPIPPPPIPPPIPPPPIPPPPIPPPPIPPPPI PPPPIPPPIPPPPIPPPPPPPPPPPPPPPPPPPIPPPPIPPPIPPPPPPP PPI

In DVD2AVI the file is 29.97 NTSC. The specs for the file suggest it should be Sequence Format Interlaced Field Order A (top first).

Isn't interlaced video every other frame has black lines? If it was telicined it should be 3 progressive then 2 interlaced?

What is the above pattern describing?


Grantman

Hi Grantman,

You're looking at the GOP ( Group of Pictures ) pattern. Not at interlaced, progressive frames. :)
The strange thing is that I see no "B" frames at all :!:
Only "I" and "P" frames.
So it's an IPPPP pattern only.

-kwag

Sorry for the confusion...
 

I understood what you wrote before.

I am just using P and I to represent progressive and Interlaced.

I am actually moving frame by frame through the video in VirtualDub and logging what I am seeing.

P = a frame without black lines. It appears good quality and solid.
I = a frame with very visible black interlace lines

There are no inbetween frames.

The P and I are just to make it easy to write out the pattern.

I assure you it is not the GOP (which is the mistake I first made trying to see if files were interlaced).

Does that make sense?

Grantman