Getting rid of interlace artifacts with HD footage

MTuggy wrote on 12/13/2011, 1:55 PM

Most of the footage I shoot is 1080p now, though some projects have a blend of 1080p footage with 1080i footage from various cameras, so I usually make the project settings at 1920x1080p.

Trouble is, I see some interlace artifacts in some of the video segments when I render to DVD quality. I don't see any issues when I render out to .m2v full HD video streams.

My question is - how can I best avoid these issues (other than using the "Reduce Interlace Flicker" which is OK, but not perfect)? Would it make any difference by how I set up the project settings or should I render out to progessive format mpg's or keep them interlaced? I experimented a bit and it feels like with a mixture of camera footage, rendering to progessive scan mpg's works better.

I don't really want to have to convert all my files to mxf or avi's intermediates but is that the solution?

Mike

Comments

farss wrote on 12/13/2011, 1:59 PM

"I don't really want to have to convert all my files to mxf or avi's intermediates but is that the solution?"

That's most unlikely to sove anything.

Have you specified a De-interlace method in you projects settings. I've found Interpolate works best.

Bob.

amendegw wrote on 12/13/2011, 2:03 PM

"Have you specified a De-interlace method in you projects settings. I've found Interpolate works best"Sage advice! Also make sure your rendering quality is set to "Best".

Edit: One more thing I've just been experimenting with is... when shooting in 1080p, you can render to 720x480 23.976 2-3 Pulldown. If you don't have a whole lot of motion, my first experiments seem to produce a very good result.

Edit2: Here's a re-shoot of the HulaDancer clip in 1080 60p, rendered directy to the "DVD Architect 24p NTSC Widescreen" template. No "Reduce Interlace Flicker", no Sharpen FX, no nothing. HulaDancer24p.zip - Not bad, eh?

...Jerry

farss wrote on 12/13/2011, 3:23 PM

"Edit2: Here's a re-shoot of the HulaDancer clip in 1080 60p, rendered directy to the "DVD Architect 24p NTSC Widescreen" template. No "Reduce Interlace Flicker", no Sharpen FX, no nothing. HulaDancer24p.zip - Not bad, eh?"

It certainly looks better. Did you shoot this exactly the same though, I cannot find the other tests you did to check this.
One of the problems with this problem is the size of the offending pattern as that determines the spatial frequency. If you want to run tests to compare cameras and recording methods I'd suggest using Zone Plate test patterns, then you know you are only dealing with one variable.

There's a free zone plate generator here. I've had a brief play around with and so far I'm not all that happy with it as it has a lot of jaggies and no anti-aliasing has been applied so you could end up looking at errors in the chart rather than the camera.

Alan Roberts of BBC engineering fame has some good info on how to use these kind of charts here. He's also selling printed charts for GBP 35 which is an absolute bargain compared to some of the prices I've found for these kinds of charts.

Bob.

amendegw wrote on 12/13/2011, 4:23 PM

"Did you shoot this exactly the same though, I cannot find the other tests you did to check this.Bob, You make a good point. Actually, I did it the reverse way - since I knew I couldn't repo the orginial 1080i footage, I re-shot the scene in with the same camera, same lighting, same everything except this time in the original format of 1080i.

Here's the render of 1080i -> "DVD Architect NTSC Widescreen video stream" template (i.e. 29.97 fps, 60i): HulaDancerFrom60i.zip Looks better than the original Hula Dancer clip, but not as good as the 24p test.

...Jerry

farss wrote on 12/14/2011, 5:43 AM

"Looks better than the original Hula Dancer clip, but not as good as the 24p test."

Interesting, veeeery interesting.
I think it confirms what I thought was happening with the 60p>24p processing. What's happening there is Vegas's frame interpolation is blending frames and blurring the parts that are moving. You're throwing away temporal resolution and reducing the spatial resolution of anything that moves. What shutter speed were you using.

Bob.

amendegw wrote on 12/14/2011, 6:24 AM

"What shutter speed were you using."I took these clips some months ago. I'm 99% sure both were both shot at 1/60 sec.

It's a pretty easy test to recreate, so I could recreate it with specific settings (although my wife might have other holiday plans for me).

...Jerry