How can I create 50p footage from 50i footage ???

Marcello Mazzilli · October 19th, 2009, 09:04 AM

Thanks for your (long) answer. I'm sure this software can help... I ran accross Virtual Dub and also FFMPEG (which I think is at the core of all these) .. There are two problems though... 1st.. I am a Mac user and all this stuff is in a PC environnement ... 2nd .. The only script I remember in PC is "format C:" in MSDos :)

Robert Martens · October 19th, 2009, 09:11 AM

Ah, I see. I haven't used a Mac since OS 8, so I'm afraid I can't offer any help there. Sorry to bother you with all that! At least my fingers got some exercise.

Bill Davis · October 19th, 2009, 06:50 PM

Sorry Marcello, but I STILL think you're doomed to poor results.

The temporal offset takes place <between the fields> NOT <between the frames>

Just do a simple still frame capture from interlaced video content.

See mine attached. SD video frame captured with both fields intact. Then enlarged slightly to make it easier to see edge detail. Click it to full size then look at the kids cheek or hairline.

You'll see that the edges of most objects in motion have a pronounced STAIR STEP appearance.

This is the temporal offset between the two fields.
The even scan started at one position - the odd scan clearly starter later.

Note, this was a shot of a kid taking off an outlet cover before painting a room - and the "motion" is just his natural body movement, not something like a car speeding by!

And temporal resolution is critical to achieving overall higher resolution.

A good example is the classic film of a trotting horse to settle a bet as to whether there was any point where the horse had all four hooves off the ground.

If your temporal resolution is not sufficient to captured enough slices of time to see a moment when all four hooves might be airborn - or at least provide enough motion and vector data to properly interpolate all possible positions of all hoofs - there would be be no way to definitively interpolate that answer.

So no matter how many still shots you take at say - 2 second intervals - unless you get really lucky, you might NEVER see the horse in that position. So you're screwed in the interpolation game.

Another good example happens all the time in security cam footage. The lower the frame rate, the lower the chance you'll get a shot of the criminal's face. And if you don't capture that data, or at least enough to make quality interpolation possible - then all the number crunching in the world won't get you a decent still of the perp.

There's nothing wrong with the idea of trying to up-rez footage. Just as long as you understand the limitations.

The computer can tell you where the passing race car might have generally been positioned between frame 100 and frame 101. But it will NEVER tell you if the driver blinked during at the time of frame 101A. And that means you're NOT increasing temporal resolution no matter how much sophisticated math you bring to bear.

Marcello Mazzilli · October 20th, 2009, 01:29 AM

That is exactly what I am saying... The two fields represent two different moments in time.... T1 and T2 that is 1/50th of a second afterwards. What I am saying is that I can pair T1 and T2 but also T2 and T3 and not only T1 and T2 and T3 with T4.

Bill Davis · October 20th, 2009, 12:48 PM

Yes, you can "pair" them. But the result will be a FUZZY frame with poor resolution since you're combining image data with DIFFERENT temporal states. Unless you somehow "line up" the two fields - and if you do that - you're NOT generating new temporal data - just estimating it - or picking one field's state and making the other match.

Either way, you get no additional ACTUAL resolution.

Marcello Mazzilli · October 20th, 2009, 01:27 PM

But they are ALL in different temporal states.. field 2 is 1/50th sec after field 1 and field 3 is 1/50th after field 2.. there is exactly the same jump between ALL the fields.. that is 1/50th of a second... When standard interpolation (from 50i to 25p) pairs 1 with 2 and 3 with 4 is just creating a new inbetween temporal state every 1/25th but there is absolutely no difference with what I want to do... Also I want to pair 2 different time states also 1/50th of a second of time distance in between

Craig Parkes · October 20th, 2009, 09:59 PM

Marcello, sounds like you need to find a programmer/scripter with experience in Avi Synth to help you out. I don't think anyone here has that experience, perhaps you should look for a forum which specializes in Avi Synth.

Paul Doherty · October 20th, 2009, 11:47 PM

Marcello if you do go to another forum and get the solution you want then please come back and post it here.

My understanding is that if you grab a single frame from 50i footage and take it into a program like Photoshop and deinterlace it then you can produce a single frame from either the odd field or the even field. So you can produce two different images 1/50th of a second apart from one frame.

So if you deinterlace 1 second of 50i footage in Photoshop and process both the odd and the even fields then you end up with 50 frames (not fields) each of which was captured 1/50th of a second apart. Yes the resolution takes a hit because you are interpolating to fill the gaps, but you have to interpolate to fill the gaps if you are going from 50i to 25p.

Anyway that I think is roughly what you want to achieve, so I'm guessing that it may well be possible to do it in a neater, more automated way using either AviSynth or After Effects.

Graham Hickling · October 21st, 2009, 01:31 AM

I apologise for giving up reading the posts in between, but to answer the OP, I recommend the TDeint plugin for avisynth, used in combination with virtualdub. Available here: http://web.missouri.edu/~kes25c/TDeintv11.zip

Write an avs text file containing something along the lines of the the following script and load it into Virtualdub, then save to a new avi file.

AviSource = "Filename.avi"
TDeint(x)

If x is set to 0 you will get 25P footage via smart deinterlace. If it is set to 1 you will get 50P footage (not just repeated frames). Its not the absolute best avisynth smart deinterlacer, but I consider it a great compromise between speed, quality amd ease of use.

October 19th, 2009, 06:50 PM	#18
Bill Davis Trustee Join Date: Jan 2004 Location: Scottsdale, AZ 85260 Posts: 1,538	Sorry Marcello, but I STILL think you're doomed to poor results. The temporal offset takes place <between the fields> NOT <between the frames> Just do a simple still frame capture from interlaced video content. See mine attached. SD video frame captured with both fields intact. Then enlarged slightly to make it easier to see edge detail. Click it to full size then look at the kids cheek or hairline. You'll see that the edges of most objects in motion have a pronounced STAIR STEP appearance. This is the temporal offset between the two fields. The even scan started at one position - the odd scan clearly starter later. Note, this was a shot of a kid taking off an outlet cover before painting a room - and the "motion" is just his natural body movement, not something like a car speeding by! And temporal resolution is critical to achieving overall higher resolution. A good example is the classic film of a trotting horse to settle a bet as to whether there was any point where the horse had all four hooves off the ground. If your temporal resolution is not sufficient to captured enough slices of time to see a moment when all four hooves might be airborn - or at least provide enough motion and vector data to properly interpolate all possible positions of all hoofs - there would be be no way to definitively interpolate that answer. So no matter how many still shots you take at say - 2 second intervals - unless you get really lucky, you might NEVER see the horse in that position. So you're screwed in the interpolation game. Another good example happens all the time in security cam footage. The lower the frame rate, the lower the chance you'll get a shot of the criminal's face. And if you don't capture that data, or at least enough to make quality interpolation possible - then all the number crunching in the world won't get you a decent still of the perp. There's nothing wrong with the idea of trying to up-rez footage. Just as long as you understand the limitations. The computer can tell you where the passing race car might have generally been positioned between frame 100 and frame 101. But it will NEVER tell you if the driver blinked during at the time of frame 101A. And that means you're NOT increasing temporal resolution no matter how much sophisticated math you bring to bear. Attached Thumbnails

October 20th, 2009, 09:59 PM	#22
Craig Parkes Major Player Join Date: Dec 2006 Location: Auckland, New Zealand Posts: 444	Marcello, sounds like you need to find a programmer/scripter with experience in Avi Synth to help you out. I don't think anyone here has that experience, perhaps you should look for a forum which specializes in Avi Synth. __________________ www.afterglow.co.nz

October 19th, 2009, 09:04 AM	#16
Marcello Mazzilli Major Player Join Date: Apr 2009 Location: ITALIA Posts: 416	Thanks for your (long) answer. I'm sure this software can help... I ran accross Virtual Dub and also FFMPEG (which I think is at the core of all these) .. There are two problems though... 1st.. I am a Mac user and all this stuff is in a PC environnement ... 2nd .. The only script I remember in PC is "format C:" in MSDos :)

October 19th, 2009, 09:11 AM	#17
Robert Martens Major Player Join Date: Jun 2003 Location: Miller Place, NY Posts: 820	Ah, I see. I haven't used a Mac since OS 8, so I'm afraid I can't offer any help there. Sorry to bother you with all that! At least my fingers got some exercise.

October 20th, 2009, 01:29 AM	#19
Marcello Mazzilli Major Player Join Date: Apr 2009 Location: ITALIA Posts: 416	That is exactly what I am saying... The two fields represent two different moments in time.... T1 and T2 that is 1/50th of a second afterwards. What I am saying is that I can pair T1 and T2 but also T2 and T3 and not only T1 and T2 and T3 with T4.

October 20th, 2009, 12:48 PM	#20
Bill Davis Trustee Join Date: Jan 2004 Location: Scottsdale, AZ 85260 Posts: 1,538	Yes, you can "pair" them. But the result will be a FUZZY frame with poor resolution since you're combining image data with DIFFERENT temporal states. Unless you somehow "line up" the two fields - and if you do that - you're NOT generating new temporal data - just estimating it - or picking one field's state and making the other match. Either way, you get no additional ACTUAL resolution.

October 20th, 2009, 01:27 PM	#21
Marcello Mazzilli Major Player Join Date: Apr 2009 Location: ITALIA Posts: 416	But they are ALL in different temporal states.. field 2 is 1/50th sec after field 1 and field 3 is 1/50th after field 2.. there is exactly the same jump between ALL the fields.. that is 1/50th of a second... When standard interpolation (from 50i to 25p) pairs 1 with 2 and 3 with 4 is just creating a new inbetween temporal state every 1/25th but there is absolutely no difference with what I want to do... Also I want to pair 2 different time states also 1/50th of a second of time distance in between

October 20th, 2009, 11:47 PM	#23
Paul Doherty Regular Crew Join Date: May 2003 Location: Sherburn, England Posts: 136	Marcello if you do go to another forum and get the solution you want then please come back and post it here. My understanding is that if you grab a single frame from 50i footage and take it into a program like Photoshop and deinterlace it then you can produce a single frame from either the odd field or the even field. So you can produce two different images 1/50th of a second apart from one frame. So if you deinterlace 1 second of 50i footage in Photoshop and process both the odd and the even fields then you end up with 50 frames (not fields) each of which was captured 1/50th of a second apart. Yes the resolution takes a hit because you are interpolating to fill the gaps, but you have to interpolate to fill the gaps if you are going from 50i to 25p. Anyway that I think is roughly what you want to achieve, so I'm guessing that it may well be possible to do it in a neater, more automated way using either AviSynth or After Effects.

October 21st, 2009, 01:31 AM	#24
Graham Hickling Trustee Join Date: May 2004 Location: Knoxville, Tennessee Posts: 1,669	I apologise for giving up reading the posts in between, but to answer the OP, I recommend the TDeint plugin for avisynth, used in combination with virtualdub. Available here: http://web.missouri.edu/~kes25c/TDeintv11.zip Write an avs text file containing something along the lines of the the following script and load it into Virtualdub, then save to a new avi file. AviSource = "Filename.avi" TDeint(x) If x is set to 0 you will get 25P footage via smart deinterlace. If it is set to 1 you will get 50P footage (not just repeated frames). Its not the absolute best avisynth smart deinterlacer, but I consider it a great compromise between speed, quality amd ease of use.