Sharp tips for 1080i HD to NTSC Wide DVD to HD?

Andy_L wrote on 6/9/2010, 5:21 PM

Anyone willing to share unsharp mask secrets for 1080i HD video rendered to 60i DVD widescreen for display on your standard modern dvd player, which upsamples back to 1080?

I've been trying really small radius settings (.5) and modest sharpening amounts (170-220), but the result I'm getting has a harsh, jittery look about it. Before I embark on another round of testing, thought I'd solicit advice.

Seems like the move from square pixels to 1.2 really goofs up the sharpness/picture quality.

Comments

John_Cline wrote on 6/9/2010, 6:40 PM

You're expecting 1920x1080 HD video which gets rescaled to 720x480 and then back to 1920x1080 to look as good as the original HD video? This is expecting WAY too much. There is absolutely no way that it's going to be as sharp as the original footage. Once the video goes from 1920x1080 to 720x480, the damage is done.

farss wrote on 6/9/2010, 7:00 PM

The critical issue is where in the pipeline you apply the sharpening. If you apply it to the HD images before downscaling all manner of issues could very well bite you, hard.

Try this. Edit on a HD timeline. If Interlaced footage make certain you specify a de-interlace method, Interpolate is generall the best. Set Project to Best in project properties.

Now Nest that project into a SD project same as your final render. Apply any sharpening in this project.

As John rightly says of course, it'll never look as good as real HD but it should not look outright woefull either. I've noticed a dramatic difference in how various upscaled SD looks in various HDTVs. You can twiddle your SD to make it friendlier to the upscalers and the horrors they can cause.
Bob.

Andy_L wrote on 6/9/2010, 8:27 PM

Bob, I bet that was the issue -- I was sharpening individual tracks, which means I was applying sharpening pre downrez. Thanks for the help!

amendegw wrote on 6/12/2010, 5:31 AM

"Try this. Edit on a HD timeline. If Interlaced footage make certain you specify a de-interlace method, Interpolate is generall the best. Set Project to Best in project properties.Two questions, using this method,

1) When I nest my HD project, my markers do not seem to be included in the import. Is there a way to fix this?
2) In order to reduce flicker, I like to add a touch of Vertical Gaussian Blur, should this be added before or after the Sharpen FX (or does it make a difference?)

...Jerry

farss wrote on 6/12/2010, 6:08 AM

1) Give me 12 hours to answer that if no one else does.

2) The vertical GB has to be added to the HD before downscaling. If adding sharpening that should be done after downconversion.
So you add the GB to the child HD project and the sharpen to the master SD project.

Bob.

craftech wrote on 6/12/2010, 6:18 AM

My experience has been that I get better results shooting with a good SD camera in the first place. It up-samples just fine.

John

farss wrote on 6/12/2010, 7:01 PM

I checked this out for you.
I created project "Child" and added two markers to the T/L.
I bought that project into project "Parent" and the makers showed up inside that event from Child. I was able to run the Promote Media Markers script (Under Tools>Scripting) and have them as timeline markers.
You might not be seeing the Media Markers because of how you have Vegas setup. Look under View>Event Media Markers. There's two switches there to enable/disable display of Media Markers and their labels.

Bob.

farss wrote on 6/12/2010, 7:08 PM

It seems to me to depend on many factors.
I only have ready access to one upscaling HDTV, a now quite old Bravia. The images from the PD170 via composite look attrocious on it. The best I've seen come from the SD composite out of the Z1, even better than BBC Digibeta played out of a J30 over composite.

I'm aware of the issues you've had with your EX1. I've bought a couple of lights (red and blue PARs) to get to the bottom of your problem. Unfortunately the task fell off my todo list when the lights got packed away :)

Bob.

Andy_L wrote on 6/13/2010, 5:36 PM

Bob, can you think of a simple experiment which would confirm whether or not an fx is applied pre or post sizing? There are a couple of scenarios I'd like to check out, and I can't think of any easy test that would be obvious.

farss wrote on 6/13/2010, 10:08 PM

Off the top of my head no. I've found some FXs that I use seem to fail to obey that pre/post triangle at the start of the FX timeline which is why I just nest projects where this is important.

If you create say a HD project then everything in that project is done at project resolution. The output of that project is then fed into the parent project and the parent does the scaling.

Bob.