Will software deinterlacing (and scaling) be better?
 

I have this old Lumagen VisionHDQ Video Processor in my home theater rack that I use to scale SD material to my projector. I have not used it for years and I figure i’ll use it to deinterlace and upscale videos when capturing. I know interlaced source should always remain interlaced and it will never look good when upscaled to HD. But since my capture card (Blackmagic Intensity Pro 4k) does not support 480p recording, I have to let the HDQ upscale it to 720 progressive. I can also capture at 1080p, but I feel it's just overkill. What i’m really after is the deinterlacing of the source material, and in my opinion, the Lumagen HDQ is one of the best hardware deinterlacer that can handle the job.

Now my question is, will software deinterlacing be better? I have tried numerous plug-ins and software (BorisFX, Apple Compressor, Yadif). I've yet to try Virtualdub/AVISynth though as it has a very steep learning curve. I'm wondering if anybody is willing to try as an experiment?

I’m providing the uncompressed NTSC file for anybody that wants to give it a go. I’m also providing a sample of what the Lumagen can do.

Equipment chain:
JVC HR-S7600U s-vhs — AVT-8710 TBC — S-video in to Lumagen VisionHDQ — DVI/HDMI output to BMI Intensity Pro 4k HDMI input.

Capture software:
Blackmagic Media Express

Samples:

4:3 Film Source (The Big Blue VHS)
Original 480i NTSC_Film Source
Lumagen 720p Upscaled_Film Source

4:3 Video Source (Vestax Turntablism VHS)
Original 480i NTSC_Video Source
Lumagen 720p Upscaled_Video Source

You might have to wait a while for readers willing to download 8.5 GB of video. 809MB of 480i is as far as I'm willing to go. Meanwhile you should know that your movie source isn't interlaced. Deintelacing is ruining your movies. The field order of the NTSC originals is reversed.

Just capture interlaced. Honestly, its easier in the end. The Intensity Pro should support 480p capture via the component or HDMI inputs. If it doesn't, typical Blackmagic crippling their hardware:P

The originals are hard-telecined.

Fair enough. The capture was from a vhs copy, Maybe it's the wrong choice of source then. I plan to capture home videos with the hardware scaler with the intention of watching it on a front projector. That's why I would like to deinterlace from the get go instead of doing it in software. I did already archive them as interlaced though.

Nah... can't capture in 480p. Believe me, if it could, I would. Tried using component output too and 480p is not an option. Yes, I did already capture the originals as interlaced. Just want to test this method and whether the result would be acceptable. I appreciate your reply.

I don't think you understand. Telecine is not interlaced. It's progressive video with 3:2 pulldown, meaning that each group of 5 frames has 3 progressive frames and 2 frames that look interlaced. Deinterlacing telecined video gives you 60fps with duplicates of every frame. The capture should be made as-is (interlaced, to you), then inverse telecined for 23.976 original film speed playback with no duplicate frames. Deinterlacing telecined video generates chroma ghosting and other problems.

And, yes, upsampling and denosing are both better in software with Avisynth, which you'll need anyway for inverse telecine.

Why is your frame size 720x486? That's not NTSC. NTSC is 720x480. Your video is Aja codec with YUVY color, which many readers will problems with. It's also soft from aggressive denosing, and color is a mess.

I'll get up more detail a little later today.

Attached are two versions of your FilmNTSC.avi, which ZI called "Film1" in my workfiles.

The attached Film1_720x480p.mp4 is 23.9786 inverse telecined progressive NTSC. It's an MP4 because 23.97 is not valid for DVD or standard definition BluRay/AVCHD. I think you'll see that with proper processing (not deinterlacing) and cleanup there would be little need for upsampling -- let your playback system do that. Your playback hardware can upsample properly to fit your display. I wouldn't resize a video without removing telecine effects and cleaning up the godawful noise, shimmer, and bad color. Cleanup of this video took 4 hours. If you restore 3:2 pulldown during encoding it would be valid for DVD or SD BluRay.

However, since you ask whether software processing would make better upsampling, the attached Film1_720p_BDAV_m2t.zip might answer your question. It's a zip file because m2t isn't an allowed forum format. The 720x486 sample was first cropped to NTSC 720x480 by removing bottom border head switching noise. After scene-by-scene cleanup and color correction (there are 11 scenes in this video) in Avisynth and Virtualdub, the video was resized to 960x720 using Avisynth's 16-bit dither functions with Spline36Resize. Black border pixels were added to each side to make the frame size valid for BluRay. The video is encoded for 720p BDAV at 23.976fps.

The Avisynth code for the 16-bit resize:
The video had a lot of noise, oversaturation, red smear. and color casts. I'm not in love with what some JVC players do to VHS color, so I did what I could given the time. Some shots still have that Digipure yellow cast and could stand more tweaking.

Moderator: if you prefer that I mount the two samples in MediaFire, I have a pro account that requires no extra clicks and no popups or ads.

External sites are not trusted. Too many have disappeared over the years, or relied on users to upkeep space that's now missing. VH and Doom9 are a mess for that reason: too many helpful threads are now UNhelpful because the samples/images are missing. That sucks. So always attach here.

- On the other hand, we want clips -- not lots of lots of huge files. Long files serve little purpose.
- But we upped the attachment limit to 99mb, lifted user quotas, and invested in a larger server in 2015 for a reason. :wink2:
- Site Staff won't download from Mediafire. That place has had too many malware problems.
- Err on the side of "attach it". :cool:

1. Never do anything out-of-hardware "while capturing". Do it after, post-capture.
2. Never deinterlace unless it's for streaming web viewing.
3. Never upscale unless it's a mixed-source project. Even then, there are caveats.

Playback hardware deinterlace will always be superior to encoding software, and is always ahead of software. That probably will not change. Even playback software deinterlace (example: VLC Yadifx2) is better than encoding software. And for the purpose of this discussion, Avisynth is considered "encoding software".

You should have a look at the O.P.'s 720p "deinterlaced" version of the Film1 clip. The playback of this huge file from deinterlacing a telecined video during capture is repeated full frames, according to this pattern:
AAABBCCCDDEEEFFGGGHHIIIJJKKK
that's 28 frames, 18 of which are duplicates, some are triplicate. The file size is 4.385 GB, over 60% of which are duplicate and triplicate frames at 59.94fps. On top of that, defects like tape noise, ringing, halos, chroma smear, etc., are accentuated. I think I'd rather capture as-is and not trust the card or its software for that kind of work.

I recompressed the original 4.35GB 720p file to Lagarith YUY2 and get 792MB, including duplicates and triplicates. The lossless YUY2 Lagarith working file for the 720p 23.96fps mp4 version I posted contains only unique frames and is 281 MB with hard telecine removed. IMO the O.P. paid a price in inefficiency, redundancy and noise to get some convenience.

A few more stats to note, if you haven't fallen asleep yet:

Thank you for the thorough explanation. I really have to read on and research more into this. One quick question though, is this a product of the capture card and the software that captured it? or is it the product of the hardware video processor?

You miss the point. Telecine is not interlace. You can deinterlace all you want 24/7/365 until the cows come home, but not with telecined source. They are not created or processed in the same way.

Thank you. Wrong choice of source sample then. I will not be capturing movies btw. I'm planning to do this with old home videos. I still have much to learn and I appreciate the knowledge that you share. But I wish you could tone down the condescending replies. I'll take down the samples.

Great advice thank you. That is what i'll be doing and do a separate capture with the hardware as an experiment.

We can close this thread now. Appreciate all that had replied.

Apologize if you mistook repetition for condescension. This forum has always been, in my experience, a source for high quality production. High quality is no longer a standard in our visual culture, where lowering the bar is making these forums a waste a time for many of us. The longer you work at your prized videos, the better you'll see. If they're not important, you can save yourself a great deal of time, trouble, and expense.