Capture card causing jitter and shifting?

[sanCapture]

After spending/wasting 2 weekends capturing some VHS tapes to HDD, I noticed that on all my captures, the bottom at top of the frame jitter back of forth noticeably. I was too busy before trying to peg down the whole 16-236 color stuff, which I still have a hard time adhering to because on nearly every program I use to view video on my computer, making videos 16-235 make them look grey washed out.

Before investing in a TBC I happened to plug in my old AverMedia UltraTV PCI 300 card to capture the h-sync line to see what it's doing. To my surprise, there was nearly no noticeable jitter in the line or video, but there is much more picture noise.

The capture card I mainly use is the WinTV-HVR-1250 PCI-E, that has much less noise. Although, the card seems to be be curving the video by about 5 pixels total and jitters at the top and bottom from my VCR or an old 8mm camcorder. It seems to be be ok though on the blank blue vcr signal.

I'm tempted with buying the AVT-8710 TBC, but with reports of high defect rates, larger dropped frames, and without knowing if it might actually be my capture card I'm a bit reluctant to get one.

[volksjager]

you really need a TBC to capture vhs
if you are concern about getting a bad 8710 either look for one of the old green colored ones or get a Datavideo TBC-1000.

what are you using for a VCR?

[lordsmurf]

Quote:

Originally Posted by volksjager

get a Datavideo TBC-1000.

Or TBC-100 (PCI card version).

[sanCapture]

Thanks for the replies, I ended up ordering the AVT-871. I hope it can make much better captures but I have my doubts.

While waiting for it to arrive I experimented with trying to use software to to correct the jitter. Most of the AviSynth scripts I found do a poor job though.

I was able to get pretty good results with a modified version of vcmohan's de-jitter plug-in that I made. It seems to work in about 95% of my captured videos. I adapted the plug-in to do the following:

Edge Detection:
1. Do standard edge threshold detection.
2. Take a weighted average of all the preceding pixel's luma.
3. Average shift amounts from x# of lines before & after.
4. Results is a floating point shift number such as 1.12345.

Shifting:
Take proportional ratios of adjacent "pixels" from shift distance.
For Example if shift is 1.75: px1 = 0.25*px2 + 0.75*px3

I find this method produced much more smooth shifting without any jarring shifts or false positives by noise.

Attached is a 2x magnification of the bottom of a VHS capture before and after my filter. maxDist: 10, thresh: 30, Avg: 3, Padding: 40, blend: true