What's going on with the fields in this source?

Inspector.Gadget · 17th September 2009, 18:17

The m2v linked below is taken from the "Arriving Somewhere" DVD.

http://www.mediafire.com/file/ydn0di..._1.demuxed.m2v

Here's the deal: originally, the source seems to have been field-based. Before authoring, it was deinterlaced to NTSC progressive (30p). If it makes any difference, the DVD is authored as NTSC interlaced (60i). No pulldown is used. The fields used to assembled each frame seem slightly off in what looks to be a temporal fashion: there are no mismatched fields across frames, but the two fields in each frame are distinct. This shows up as some sort of texture throughout the frame, rather than the classic aliasing "stairstep" on edges. If it's relevant, the system MeGUI uses to check the source type reports "Progressive".

I know the content isn't interlaced. Since it was deinterlaced before encoding, is it likely to be have phase-shifted fields or merely be aliased or what? Thanks in advance for the help!

thewebchat · 17th September 2009, 23:30

These are just deinterlacing artifacts. Use a blur filter like EEDI2 or SangNom to smooth them out.

Inspector.Gadget · 17th September 2009, 23:57

Thanks for the advice. Won't doing that discard one field though?

thewebchat · 18th September 2009, 04:59

Yes, but that field was guessed by deinterlacing anyway so you're not missing out on much. Figure out if the top or the bottom field is the original and interpolate the other. Use half-rate deinterlacing since half the fields are redundant.

Inspector.Gadget · 18th September 2009, 05:25

Ah, I see. That makes sense. Thanks!

17th September 2009, 18:17	#1 \| Link
Inspector.Gadget Registered User Join Date: May 2008 Posts: 1,618	What's going on with the fields in this source? The m2v linked below is taken from the "Arriving Somewhere" DVD. http://www.mediafire.com/file/ydn0di..._1.demuxed.m2v Here's the deal: originally, the source seems to have been field-based. Before authoring, it was deinterlaced to NTSC progressive (30p). If it makes any difference, the DVD is authored as NTSC interlaced (60i). No pulldown is used. The fields used to assembled each frame seem slightly off in what looks to be a temporal fashion: there are no mismatched fields across frames, but the two fields in each frame are distinct. This shows up as some sort of texture throughout the frame, rather than the classic aliasing "stairstep" on edges. If it's relevant, the system MeGUI uses to check the source type reports "Progressive". I know the content isn't interlaced. Since it was deinterlaced before encoding, is it likely to be have phase-shifted fields or merely be aliased or what? Thanks in advance for the help! Last edited by Inspector.Gadget; 17th September 2009 at 20:57.

17th September 2009, 23:30	#2 \| Link
thewebchat Advanced Blogging Join Date: May 2009 Posts: 480	These are just deinterlacing artifacts. Use a blur filter like EEDI2 or SangNom to smooth them out.

17th September 2009, 23:57	#3 \| Link
Inspector.Gadget Registered User Join Date: May 2008 Posts: 1,618	Thanks for the advice. Won't doing that discard one field though?

18th September 2009, 04:59	#4 \| Link
thewebchat Advanced Blogging Join Date: May 2009 Posts: 480	Yes, but that field was guessed by deinterlacing anyway so you're not missing out on much. Figure out if the top or the bottom field is the original and interpolate the other. Use half-rate deinterlacing since half the fields are redundant.

18th September 2009, 05:25	#5 \| Link
Inspector.Gadget Registered User Join Date: May 2008 Posts: 1,618	Ah, I see. That makes sense. Thanks!