Really? - To me the alpary version looks a bit more "puzzled" around the window. Just like artefacts.

Marco

"Looks like you only used the interpolate setting in Vegas. I think the blend setting would give you results just like DV-Filmmaker.

Marco"

Uhm, the results in Vegas are better than those of DVF.

I see the same thing with the Vegas interpolate method, but the
alparysoft "simple interpolation" method renders cleaner lines
and a brighter and whiter looking chrome around the car window.

Ah, yes, I looked at the wrong alpary version. You are right with the simple interpolation mode.

Marco

Oh, you meant it this way. Yes.

During my own tests in the past I found it is very difficult to find a "winner tool" or "winner method" for a perfect deinterlacing. Results differ so much dependend of the pictures and motion in the video. In some cases Vegas interpolation seems to be perfect beccause of the sharpness, sometimes blending is better because of the smoothness of motion. Sometimes having two tracks with different field order blended into each other is better. And sometimes external tools do a better job. And sometimes even the "Reduce Interlace Flicker" function isn't that bad, though it does only blur the fields a bit.
So one might analyse each and any event and choose the best individual method which might take a very long time. But at least I find the Vegas internal ways to manage deinterlacing are rather good and before spending money for maybe expensive tools I'd took one of the cheaper ones ;-)

Marco

Hey there,
I've been playing with PAL to NTSC conversions recently, and one of the first steps is deinterlacing the PAL footage. There were considerable differences that I saw on how the deinterlacing process was achieved, and subsequently resulting in varying quality.
With the converting of PAL to NTSC much depended on actually watching the footage back through the VEGAS timeline, in order to tell which one yielded better results. Just looking at one still frame told little about how the actual footage appeared once viewed on a telly or monitor.
I found this program to do a great job: DVfilm Atlantis. They also have a toned down program (albeit still a hundred bucks), called "Maker". For kicks, you could try their free download (puts their logo on the footage). It's at:
http://www.dvfilm.com/maker/index.htm
The program requires QuickTime, and a version not later than 6 (6.5 apparently causes probs). They recommend version 5.
Marcus is claiming that his deinterlacing process can just about loose no resolution, something which I found to be about true using my conversions using Atlantis.

Cheers,
MAL
--------------

Hi Mal, different places - same guys ;-)
I'll give DVfilm Atlantis a try. I'll gonna do some further deinterlacing tests in the next days because somebody told me, best deinterlacer for owner of the XL-1 would be the XL-1. Means he preferes using the frame mode of this camera to avoid prostpro deinterlacing. But I am convinced postpro deinterlacing is the better way to go.

Marco

This has got me thinking. For material that I know beforehand is going to be converted to NTSC best thing would be to shoot 25p, convert to 24p and then pulldown to 29.97 NTSC or if it's on DVD let the player do the pulldown.

Also anyone know if any of the above mentioned products use motion vectors? That does seem to be the only way to get really good results. I know there's expensive hardware boxes that do it, surely someone provides a cheaper way to do it in software.

Heya Marco! ;-)))

According to writings (including what's onthe DVfilm website) the XL1 uses a quasi-progressive mode, and DOES loose resolution when it "deinterlaces" in its frame mode (360 lines vs. 330 lines IIRC). So yeah, the better resolution should be obtainable using DVfilm's MAKER.

Cheers,
MAL
---------

farss,

motion vectors are a tricky beast. With all what I've seen (and those are only the "cheaper" ones) it rarely offers the best quality. Like Marco mentioned further up though, it all depends on the footage.

Cheers,
MAL
-----------

Yes, XL-1 frame mode lose resolution. But I just made a test now shooted some footage with my XL-1 used the normal interlace mode and the frame mode with equal camera motion.

Then I took the interlaced and "progressive" footage into Vegas.
I used Vegas deinterlacing once in blend and once in interpolate mode. And I tried using that method of overlaying the clip twice having one set to upper fields first with 50 percent opacity.

None of the Vegas internal method was as good in quality as the XL-1 framemode was. The XL-1 picture is visible sharper even compared to Vegas interpolat method and motion keeps being rather smooth (though you cannot preserve the smoothness of interlaced footage).

Now I downloaded and installed the DVfilm tool. Testing ...

Marco

I had a chance last night to fool around with some of these tools.

From what I can tell, the latest free deinterlacer in VirtualDub by Donald Graft (version 2.7) gives me the cleanest results. He separates moving from static areas. The static areas get interlaced, with full vertical resolution. The moving areas get blended using cubic interpolation.

There is also an option to show just what areas will be left alone and what areas will be interpolated. Neat, and it's free.

http://neuron2.net/smart.html

Yes, my results with the Smart Deinterlacer for VD are pretty good too. But still the XL-1 frame mode is a touch better.

I'm still fighting with the dvFilm tool. Till now it doesn't completely remove the interlace traces of my footage. But maybe I simply missed the best settings. Still testing ...

Marco

O.k. - I think I can't get more out of those tools. To me VirtualDub Smart Deinterlacer gives me almost same results as DVfilm maker. Good points about Smart Deinterlacer are that it can be used inside Vegas through the DV-plugin and it for free.

But none of the tools results look as good as the XL-1 frame mode. I'll post some pics later.

Marco

Good one, Marco. Would like to see the pics.
Maybe I am really doing something wrong then in VEGAS, as to me the results are quite obviously better in Atlantis (for the PAL to NTSC conversion).

Hmm....

MAL
------------

Got one test uploaded now. Here I did NOT test the sharpness, but what happens to sloping edges.

Deinterlacing Test

In the pictures where I marked a roof with a red circle there you will see some aliasing.
It is quite visible in the dvfilm maker version just like as it is in the Vegas interpolate version. Just a little bit better in the VirtualDub Smartdeinterlacer version. But SmartDeinterlacer fails in reproducing some areas in the tree (also marked red).

Virtual Dub Alpari Deinterlacer did the best postpro job here. But only the XL-1 frame mode succeeded in completly removing the aliasing artefacts on the roof edges. But having aliasing artefacts within those edges will result in a slight flickering when playback the video.

Sharpness test will follow but maybe no more today.

Marco

Mmh, I made so much tests now even for the sharpness ...
... yes, dvfilm maker is rather good. I find VirtualDub Smart Deinterlacer is not worse (after I succeeded to get the damned artefacts out of the trees).

But actually none of them seem to beat the Canon's cameras frame mode. I never would have thought of this. Any software solution gives me either even a bit more unsharp picture or - it is sharp but then there are stairsteps on some kind of lines which tend to flicker on motion. Canons frame mode gives me perfect lines in any case, is not that sharp as the interlaced version but using the Vegas Unsharp Mask filter helps a lot here.

So in some cases I would prefer using my XL-1 frame mode (if I'd be ABSOLUTELY sure I wouldn't have the need for anything else but deinterlaced footage).
But if I'm forced to do a postpro deinterlacing, mmh ... I think I would use the Smart Deinterlacer.

Maybe we could share some footage to test various kind of video with different tools!?

Marco

But wouldn't you assume that frame mode was the ideal? Every scan line is present in every frame. The de-interlacers can only hope to be roughly as good if they are "smart" (i.e. areas not moving use both fields, areas moving use just one field and blur it a bit). In other words, even a smart de-interlacer will have worse vertical resolution in areas of motion. A dumb de-interlacer will have worse vertical resolution everywhere.

The only sacrifice a progressive camera has to make is that it either has to have a doubled data rate, or the temporal resolution is halved.

Almost. But Canon's frame mode is NOT a progressive mode. It loses lines of resolution - always, even in non-motion areas. The loss of resolution is about 12 percent using a frame mode.

So, yes, frame mode seems to be ideal compared to any software deinterlacing. But frame mode is not as good as real progressive modes.

Marco

How do you open the virtual dub filters using the plugnplay paK? Do you have to put the virtual dub filters in the same folder as the Vegas ones?

Okay. Good stuff!
I too have now yielded good results with SMART DEINTERLACER. Very similar to DVfilm in look. So that's certainly the way to go. As it's cheaper! ;-)
Thanks for all the help with this one!!

I haven't found a way to do the PAL/NTSC conversion as a one-step process though. I first deinterlaced, and then in a new NTSC project converted the PAL progressive footage (the deinterlaced one) to NTSC interlaced.

Cheers,
MAL
----------------