How can my capture be improved?

[Sunk]

Hello there. I go by the same name on the Digital FAQ, but have been lurking this forum as well since about early January to learn some things about transferring VHS tapes. I've decided to post here because I know that, even though many users from VH also hang out over on DigiFAQ and vice-versa, some may not.

Recently, I've made a capture of one of my tapes I've recorded in 2007. I live in Russia so our tapes are split between MESECAM and PAL, with the former being more common for TV broadcast recordings and the latter for the licensed commercial film and cartoon tapes. The capture I have was made from a MESECAM tape, using the following workflow:

JVC HR-S7600AM -> ES10 passthrough -> Matrox MXO2 LE

From what I understand, there was officially no neither line, nor frame TBC for MESECAM, so I was forced to use the ES10. Another problem is that on its own, the MXO2 LE doesn't work with MESECAM signal, meaning that, once again, I had to use the ES10 to convert the signal to PAL (which I believe is somewhat of a bad practice). I also have no standalone frame TBC yet, but I'm considering getting one if my financial situation would allow me.

Looking forward to hearing your comments, criticism and suggestions.

[jagabo]

Many things one can do to remove noise will work fine on still shots but ruin motion shots. So it's best to provide a mix of still and motion shots. Including a medium speed panning shot will be very helpful. And don't upload a sample from the intro or credits, provide something from the body of the show (intro and credits are often very different from the body). Attached is an example where I cheated -- I just averaged together all the frames. The result looks good -- but you couldn't do this with the main program.

[Sunk]

I made a new capture today from the same tape, this time where there are some camera angles and motion. I also made a capture without the ES10, just so that you could see how well the MXO2 handles MESECAM on its own (badly...)

And I do apologize for the file sizes, I just try to upload raw captures without any sort of edits. Let me know if the size can be reduced without harming the video.

[jagabo]

Maybe something like this?

LWLibavVideoSource("testcap2_es10.avi", cache=false, prefer_hw=2) 
Crop(6,0,-10,-0)
QTGMC(FPSDivisor=2)
TurnRight().QTGMC(InputType=2).TurnLeft()
SMDegrain(tr=3, thSAD=200, refinemotion=true, contrasharp=false, PreFilter=4, mode=0, truemotion=true, plane=4, chroma=false)
MergeChroma(Spline36Resize(width/2, height).aWarpSharp2(depth=20).Spline36Resize(width,height))

I didn't address the red bar at the bottom of the frame. And why are there so many macroblock artifacts? They are too sharp to be on the tape. The implication is that the video has been encoded with a lossy codec (MPEG 2?) between the capture and your huffyuv upload.

[Sunk]

Maybe something like this?

LWLibavVideoSource("testcap2_es10.avi", cache=false, prefer_hw=2) 
Crop(6,0,-10,-0)
QTGMC(FPSDivisor=2)
TurnRight().QTGMC(InputType=2).TurnLeft()
SMDegrain(tr=3, thSAD=200, refinemotion=true, contrasharp=false, PreFilter=4, mode=0, truemotion=true, plane=4, chroma=false)
MergeChroma(Spline36Resize(width/2, height).aWarpSharp2(depth=20).Spline36Resize(width,height))

I didn't address the red bar at the bottom of the frame. And why are there so many macroblock artifacts? They are too sharp to be on the tape. The implication is that the video has been encoded with a lossy codec (MPEG 2?) between the capture and your huffyuv upload.

I'm not entirely sure what you mean by macroblock artifacts. Could you show a frame from my capture where it's most prominent?

Here are my vdub settings:



And my Matrox settings:



As for the red bar, it was most likely caused by the aged tape. None of my other tapes have it and it's also present on my lower-tier LG deck, so it's not a problem with this particular JVC unit.

[jagabo]

Never mind the comment about the macroblock artifacts. It was a problem with the way I was viewing the video.

In the processed clip I uploaded earlier I kept the frame rate at 29.97 fps because the bulk of your sample had progressive frames. At the very end it switches to some live video that is probably truly interlaced -- but there's so little motion it's hard to say for sure. For interlaced video you'll want to use 59.94 fps. Just a small change to the script:

LWLibavVideoSource("testcap2_es10.avi", cache=false, prefer_hw=2) 
Crop(6,0,-10,-0)
QTGMC()
SMDegrain(tr=3, thSAD=200, refinemotion=true, contrasharp=false, PreFilter=4, mode=0, truemotion=true, plane=4, chroma=false)
MergeChroma(Spline36Resize(width/2, height).aWarpSharp2(depth=20).Spline36Resize(width,height))

That's just a start. There are lots of other things you can do.

[Sunk]

Never mind the comment about the macroblock artifacts. It was a problem with the way I was viewing the video.

In the processed clip I uploaded earlier I kept the frame rate at 29.97 fps because the bulk of your sample had progressive frames. At the very end it switches to some live video that is probably truly interlaced -- but there's so little motion it's hard to say for sure. For interlaced video you'll want to use 59.94 fps. Just a small change to the script:

LWLibavVideoSource("testcap2_es10.avi", cache=false, prefer_hw=2) 
Crop(6,0,-10,-0)
QTGMC()
SMDegrain(tr=3, thSAD=200, refinemotion=true, contrasharp=false, PreFilter=4, mode=0, truemotion=true, plane=4, chroma=false)
MergeChroma(Spline36Resize(width/2, height).aWarpSharp2(depth=20).Spline36Resize(width,height))

That's just a start. There are lots of other things you can do.

Are you sure it should be 29.97 and 59.94? From what I understand, both PAL and SECAM, including MESECAM, run at 25 fps by default, 50 when the frame rate is doubled. Your numbers seem to be standard for the NTSC signal.

[jagabo]

Yes. I was just operating from memory and forgot your source was PAL. So the frame rates are 25 and 50, not 29.97 and 59.94. Sorry for the confusion.