The infamous interlaced video (some thoughts about converting formats, take 1) at DVinfo.net

David Lorente · November 14th, 2008, 06:24 AM

I'm currently editing a video with footage coming from a XDCAM EX (clips are in 1080p25), and old Betacam SP tapes shot 10 years ago (that's 576i50, obviously) that I have digitized with a Blackmagic Multibridge in Apple ProRes HQ. The final edit will be converted to DVD to sell to the public, so I decided to make a PAL sequence in FCP. The Beta SP clips are working nicely, but the problems come when we put HD clips in the PAL sequence.

The way FCP handles the conversion to interlaced is HORRENDOUS. I can't understand why they still haven't managed to understand the way these things work (that said, Premiere doesn't work much better, and the same can be said about most NLE's). And it is so simple. In case some smart guy responsible of this is reading the forums, I will explain it, just open your ears:

1- Fields are in fact independent frames. Back in the era of vacuum tubes, when television was invented, engineers saw that resolution could be somewhat improved by using the interlacing trick, but that was just that, a trick that displaced consecutive frames by half a line vertically.

But, when things became digital, instead of conserving the independency between fields, some smart guy thought that two fields could be blended to make a full frame. And that's when the problems started.

Because for almost every operation that involves a displacement of the pixels in one or another way, the interlaced frames must be deinterlaced in two fields, processed, and then reinterlaced again. And in most cases, this is not being done properly.

2- When a progressive image has to be converted to interlaced, you can't just pick the odd lines to make the odd field, and the even lines to make the even field. NO. THAT'S NOT THE WAY VIDEO CAMERAS WORK. For God's shake, just read a good book about video cameras, and you'll learn that, in order to produce a good image, without aliasing, flickering, and all the other weird effects related with interlacing, what video cameras are doing is just BLENDING TWO CONSECUTIVE ROWS OF PIXELS IN THE CCD TO MAKE EACH VIDEO LINE. Rows 1 & 2 form line 0, rows 3 & 4 forms line 2, etc. And then, in the next field, they switch pairs. Rows 2&3 form line 1, rows 4&5 forms line 3... And this applies not only to SD cameras, but also to HD cameras.

So I just wonder WHY AREN'T NLE'S WORKING LIKE THAT. Just give it a try, put an HD clip in a SD sequence, and you'll discover a full bunch of weird effects.

3- Here is the workflow to do a proper progressive HD to interlaced SD conversion:

a- scale down the frame to the SD format of your choice.
b- generate the fields by blending couples of lines, just the way CCD cameras work

And you may be thinking: Does not Compressor convert one scanning method into another? And the answer is: NO. It just performs a resizing, but there is not any kind of processing, even forcing the settings in the "Frame controls" tab.

I think we, as users, tired of obtaining dissapointing results not by our fault, or by a bad quality footage, but by bad implementation of image processing in the NLE's, we should made pressure to software developers to solve these problems. Because I don't need more absurd special FX and transitions, or the capability to handle a hundred video tracks, or other useless things, but I ABSOLUTELY NEED TO KEEP A GOOD QUALITY IN THE IMAGE.

Bruce James · November 14th, 2008, 06:48 AM

If you captured at 25p then you have 25 full frames of motion.
If you convert that to 50i then you still only have 25 full frames of motion but over 50 fields.

Bruce James · November 14th, 2008, 07:26 AM

Theres some interesting info here FCP sees 1080p25 as Interlaced

David Lorente · November 14th, 2008, 07:37 AM

Of course, but I'm not talking about the motion. I'm talking about the resizing of the image.

Converting 25p to 50i means that you have tocreate two fields from one frame. The problem is how these fields are created. And it is easier than you think. A way to do it is, in Photoshop, resizing a 1080p frame to 576, and then, apply a "Movement blur" with 90 degrees and 1 pixel blur. I have always used this method to create stills from high resolution graphics or photos. If you just resize a high resolution image, the flickering will be so severe that the resulting video will be unwatchable in a CRT monitor. LCD/Plasma screens work in a different way and won't show this effects, but we have to think there's still a lot of CRT's in service.

However, what I'm complaining about is of aliased lines, and misalignement between both fields. NLE's such as Premiere or Final Cut, that are being sell as professional things, can't do such a bad job in resizing images.

David Lorente · November 14th, 2008, 07:47 AM

Quote:

Originally Posted by Bruce James

Theres some interesting info here FCP sees 1080p25 as Interlaced

The problem that this guy is having is that FCP "thinks" that HDV is always interlaced, and it imports it as interlaced. The problem is easily solved by opening the "item properties" of the clip, Control-clicking over "Field dominance" setting (which may be "Upper (Odd)" or "Lower (Even)") and setting it to "None".

XDCAM EX clips shot in progressive mode are set to "Field dominance: None" and I have never had this problem.

Aric Mannion · November 14th, 2008, 10:07 AM

Why are you trying to make an interlaced video? Why don't you just de-interlace your BetaCam footage and work progressively from now on? 480p media in DVD studio pro looks great.

David Lorente · November 14th, 2008, 05:02 PM

Quote:

Originally Posted by Aric Mannion

Why are you trying to make an interlaced video? Why don't you just de-interlace your BetaCam footage and work progressively from now on? 480p media in DVD studio pro looks great.

Aric, DVD's are constrained to PAL (or NTSC) system limitations, so they will ALWAYS be 50i (or 60i). And what you say might work, but it is just a workaround. The point is I just don't understand why FCP is making such a poor job when downconverting HD to SD, knowing, as I have said before, that it is not so difficult to make it well. It is just a problem of knowing EXACTLY how interlaced video works. And after what I've seen, I think that software developpers just don't know that.

By the way, though I miss some features I had in Premiere and the very old FAST VideoMachine (you may have never heard about it, but this was a NLE released in 1993 by the german company FAST, and that worked quite well for the era), I'm overall happy working with the Apple suite, but there's still a lot of things to improve in it.

Aric Mannion · November 19th, 2008, 11:55 AM

As I understand it, DVDs are interlaced. But I've read many threads on this forum that say that you can make a 480p DVD in DVD studio pro. I argued at first, until I gave it a try. 480p video imported into DVD studio pro produces beautiful DVDs without the interlacing lines you get from importing 480i.
I've tried the same thing on idvd and it interlaced my progressive video as you would expect.
I don't think it's a work around, I think it looks better and I even de-interlace with after effects before bringing the video into dvd studio pro. Final Cut does a horrible job de-interlacing, so if you have a better tool you should really try it and see the results before you write it off.

November 14th, 2008, 06:24 AM	#1
David Lorente Regular Crew Join Date: Dec 2007 Location: Cuenca (Spain) Posts: 92	The infamous interlaced video (some thoughts about converting formats, take 1) I'm currently editing a video with footage coming from a XDCAM EX (clips are in 1080p25), and old Betacam SP tapes shot 10 years ago (that's 576i50, obviously) that I have digitized with a Blackmagic Multibridge in Apple ProRes HQ. The final edit will be converted to DVD to sell to the public, so I decided to make a PAL sequence in FCP. The Beta SP clips are working nicely, but the problems come when we put HD clips in the PAL sequence. The way FCP handles the conversion to interlaced is HORRENDOUS. I can't understand why they still haven't managed to understand the way these things work (that said, Premiere doesn't work much better, and the same can be said about most NLE's). And it is so simple. In case some smart guy responsible of this is reading the forums, I will explain it, just open your ears: 1- Fields are in fact independent frames. Back in the era of vacuum tubes, when television was invented, engineers saw that resolution could be somewhat improved by using the interlacing trick, but that was just that, a trick that displaced consecutive frames by half a line vertically. But, when things became digital, instead of conserving the independency between fields, some smart guy thought that two fields could be blended to make a full frame. And that's when the problems started. Because for almost every operation that involves a displacement of the pixels in one or another way, the interlaced frames must be deinterlaced in two fields, processed, and then reinterlaced again. And in most cases, this is not being done properly. 2- When a progressive image has to be converted to interlaced, you can't just pick the odd lines to make the odd field, and the even lines to make the even field. NO. THAT'S NOT THE WAY VIDEO CAMERAS WORK. For God's shake, just read a good book about video cameras, and you'll learn that, in order to produce a good image, without aliasing, flickering, and all the other weird effects related with interlacing, what video cameras are doing is just BLENDING TWO CONSECUTIVE ROWS OF PIXELS IN THE CCD TO MAKE EACH VIDEO LINE. Rows 1 & 2 form line 0, rows 3 & 4 forms line 2, etc. And then, in the next field, they switch pairs. Rows 2&3 form line 1, rows 4&5 forms line 3... And this applies not only to SD cameras, but also to HD cameras. So I just wonder WHY AREN'T NLE'S WORKING LIKE THAT. Just give it a try, put an HD clip in a SD sequence, and you'll discover a full bunch of weird effects. 3- Here is the workflow to do a proper progressive HD to interlaced SD conversion: a- scale down the frame to the SD format of your choice. b- generate the fields by blending couples of lines, just the way CCD cameras work And you may be thinking: Does not Compressor convert one scanning method into another? And the answer is: NO. It just performs a resizing, but there is not any kind of processing, even forcing the settings in the "Frame controls" tab. I think we, as users, tired of obtaining dissapointing results not by our fault, or by a bad quality footage, but by bad implementation of image processing in the NLE's, we should made pressure to software developers to solve these problems. Because I don't need more absurd special FX and transitions, or the capability to handle a hundred video tracks, or other useless things, but I ABSOLUTELY NEED TO KEEP A GOOD QUALITY IN THE IMAGE. __________________ Wire telegraph is a kind of a very, very long cat. You pull his tail in New York and his head is meowing in Los Angeles. Radio operates exactly the same way, but there is no cat.

November 14th, 2008, 06:48 AM	#2
Bruce James New Boot Join Date: Apr 2007 Location: New Milton,UK Posts: 19	If you captured at 25p then you have 25 full frames of motion. If you convert that to 50i then you still only have 25 full frames of motion but over 50 fields. __________________ -- http://www.pomegranate.ltd.uk

November 14th, 2008, 07:26 AM	#3
Bruce James New Boot Join Date: Apr 2007 Location: New Milton,UK Posts: 19	Theres some interesting info here FCP sees 1080p25 as Interlaced __________________ -- http://www.pomegranate.ltd.uk

November 14th, 2008, 07:37 AM	#4
David Lorente Regular Crew Join Date: Dec 2007 Location: Cuenca (Spain) Posts: 92	Of course, but I'm not talking about the motion. I'm talking about the resizing of the image. Converting 25p to 50i means that you have tocreate two fields from one frame. The problem is how these fields are created. And it is easier than you think. A way to do it is, in Photoshop, resizing a 1080p frame to 576, and then, apply a "Movement blur" with 90 degrees and 1 pixel blur. I have always used this method to create stills from high resolution graphics or photos. If you just resize a high resolution image, the flickering will be so severe that the resulting video will be unwatchable in a CRT monitor. LCD/Plasma screens work in a different way and won't show this effects, but we have to think there's still a lot of CRT's in service. However, what I'm complaining about is of aliased lines, and misalignement between both fields. NLE's such as Premiere or Final Cut, that are being sell as professional things, can't do such a bad job in resizing images. __________________ Wire telegraph is a kind of a very, very long cat. You pull his tail in New York and his head is meowing in Los Angeles. Radio operates exactly the same way, but there is no cat.

November 14th, 2008, 10:07 AM	#6
Aric Mannion Major Player Join Date: Apr 2006 Location: Boston, Massachusetts Posts: 616	Why are you trying to make an interlaced video? Why don't you just de-interlace your BetaCam footage and work progressively from now on? 480p media in DVD studio pro looks great.

November 19th, 2008, 11:55 AM	#8
Aric Mannion Major Player Join Date: Apr 2006 Location: Boston, Massachusetts Posts: 616	As I understand it, DVDs are interlaced. But I've read many threads on this forum that say that you can make a 480p DVD in DVD studio pro. I argued at first, until I gave it a try. 480p video imported into DVD studio pro produces beautiful DVDs without the interlacing lines you get from importing 480i. I've tried the same thing on idvd and it interlaced my progressive video as you would expect. I don't think it's a work around, I think it looks better and I even de-interlace with after effects before bringing the video into dvd studio pro. Final Cut does a horrible job de-interlacing, so if you have a better tool you should really try it and see the results before you write it off.