Holy Cow! Deinterlace that much of a difference???

[heavyharmonies]

Maybe I just never understood what the deinterlace filter in TMPGENC did.

Scenario:

I'm capturing a 17-year-old VHS tape to DV via Canopus ADVC-100, then encoding to MPEG1 in TMPGENC - 352x240, CQ100, 12000 kbps max (this set of settings makes WONDERFUL MPEG1 vids, and Nero will allow you to use this nonstandard format in its VCD template; my DVD player plays it wonderfully. Anyway...)

The tape is a mixture of music videos and live concert footage. The videos look fine without any filters being used, but the live footage ends up with horrible horizontal lines through various portions of the video, especially noticible on lighter hues and during fast motion; very disturbing.

For giggles I turned on TMPGENC's deinterlace filter, using the "double" setting. Problem fixed!

Interesting side effect: 5:27 video interlaced: 197MB. Same clip but deinterlaced: 77MB. HUH??? Less than half the size???

Somebody please splain dis to me.

Also, why would the "interlacing artifacts" (for lack of a better term) be noticible on the live footage but not the canned music videos off the same tape?

Maybe I should start deinterlacing all my VHS captures?

[SLK001]

I think that the problem is with your 17 year-old tapes in that the sync pulses are not being detected properly by your capture. I don't think that deinterlacing is really the reason for your pickup in file space.

This is happening to me. I capture directly to MPEG2 and with a good tape (or laserdisk), audio and video are in perfect sync. However, once a substandard sync pulse in introduced, audio starts to lag behind. At the end of a 20 minute show, audio is as far as 15 seconds behind. I'm assuming that the audio is getting its timing cue from the sync pulse... each one it misses causes audio to loose 1/60 of a second (NTSC - 60 fields a second). I'm afraid that I'll have to build a TBC and SYNC restorer just to cap these videos.

[TheInformer]

First of all, let me assume you are playing your VHS tape and running the output to your ADVC-100 device. If this is so, then your captured video *is* interlaced. Interlacing is a subject I'm not touching in detail right now because of the time needed to explain it properly and my fingers are lazy. Suffice to say, interlacing is where you take two 240 vertical line fields and combine them into a 480 line frame. Deinterlacing (for what you are doing) takes the 480 line frame and combines adjacent lines to create a 240 line frame. Maybe that's where the difference comes from?

[Xesdeeni]

1. If you are turning your VHS tapes to VCD, then yes, you should deinterlace in some way. You are experiencing some of the reasons why. If you are creating SVCDs or DVDs, then you don't want to deinterlace. For NTSC, you either want to inverse the telecine, or stay interlaced. The choice depends on the material. If all of the material is from film, you should inverse the telecine. This saves about 20%, which will improve the video quality. If any of the content is video (i.e. not from film), then you want to leave the interlace.

2. The reason the live parts looked worse was because they were actually video, where each field represents a different instant of time. The canned videos are actually shot on film. So there are only 24 instants of time per second. For transmission, these 24 frames are converted to fields. However, since there are only 24 instants of time in the source material, some fields represent portions of the same instant of time. For NTSC, three successive fields will represent the same frame and therefore the same instant of time. Then the next frame will be used for the following two fields. This is the ubiquitous 3:2 pulldown. There is still some interlace artifacting visible when two successive fields are from two different frames, but this happens only 24 times per second. In true video, this happens 60 times a second, so it is much more visible.

3. The filesize difference may be due to the reduced vertical "noise" in the source after you deinterlace. MPEG uses a mathematical model for the image. The model takes pairs of fields together. If there is motion between them, there is a great deal of vertical discontinuity...what you see as a "combing" effect. The mathematical formula blows up at this high frequency component, so the compression suffers. When you deinterlace, you remove this vertical discontinuity, so it makes sense that you are seeing smaller file sizes.

Xesdeeni