Video restoration newbie: Avisynth scripts?

[latreche34]

Lordsmurf if you think it's the VCR is not good, I have a cap from a Toshiba VHS with no TBC captured with VMC-1 TBC ON in DV quality it doesn't say much but at least mechanically it gives an idea about the playback condition;

[sanlyn]

The Toshiba seems is playing the tape more smoothly.

This leads me to ask, now that you've seen some video work and details here and there, what do you think of it?

IMO play is smoother but otherwise the video looks a lot worse than earlier ones. It has severe levels cutoff, not to mention a flood of chroma noise, DV compression effects, and a very grimy look.

If no one here can determine where all this crushing and clipping are coming from, I can't continue in good conscious to recommend any USB capture device from the list referred to earlier. Other users here and in other forums are having the same problem with those devices. In this particular case it's a mystery because this video is captured as DV. So it must have entered the PC via Firewire unless Latreche34 managed to somehow reconfigure the convoluted system referred to earlier. In any case, the video has hard clipping at Y=16 to 235 and no data whatever beyond those points.

[latreche34]

Don't you think virtualDub is not capturing smoothly, I've had instances where I stop capture and start again when virtualDub start to stutter, I know with DV didn't have that problem, JVC HR-S7600 is a reputable VCR AFAIK, I may have to do a short DV capture for the same video segment with it to isolate the problem.

[msgohan]

Quote:

Originally Posted by sanlyn

In this particular case it's a mystery because this video is captured as DV. So it must have entered the PC via Firewire unless Latreche34 managed to somehow reconfigure the convoluted system referred to earlier. In any case, the video has hard clipping at Y=16 to 235 and no data whatever beyond those points.

What's your DV decoder? The clip ranges from 0-251. Here's one frame that has almost the full range of the clip.
latreche34-NightGatherDV-ColorYUVanalyze.jpg

[sanlyn]

Quote:

Originally Posted by msgohan

What's your DV decoder?

What's yours?

DVSD:


Cedocida:


Glad you chimed in. Maybe you can explain the histograms to the O.P. and other readers. Or maybe not. I gave up on analog to DV.

[msgohan]

Mine is Cedocida, in this case outputting YV12. It seems something is converting your Cedocida output to RGB before converting it back to YUV (the chroma values have been changed, in addition to the 16-235 clipping).

Any idea what the actual decoder is for your DVSD frame? "DVSD" is the FourCC for plain DV; I don't believe it identifies any particular software.

[sanlyn]

Quote:

Originally Posted by msgohan

Any idea what the actual decoder is for your DVSD frame? "DVSD" is the FourCC for plain DV; I don't believe it identifies any particular software.

ffdshow.

Cecocida DV YV12 only:


Cecocida DV YV12 only, levels TV>PC

[msgohan]

Okay, your luma matches my Cedocida output now. Your chroma values are slightly different. I can't seem to force ffdshow to decode DV, to see what I get there.

Your latest shots are from frame 123 instead of 124, BTW.

[sanlyn]

I have received this PM request from member doodal:

Quote:

I'm trying to get the script you posted here Video restoration newbie: Avisynth scripts? to work, but I keep getting "there is no function named awarpsharp2". I have awarpsharp2.dll in my plugins folder, so I'm not sure what I'm doing wrong. Do you have any idea?

Please don't PM with technical questions such as this. Other users can't share this information in PM.

I don't know of an "aWarpSharp2.dll". The plugin is awarpsharp.dll. The latest version is dated 2016-06-24. The download page is here: http://avisynth.nl/index.php/AWarpSharp2. The download has 32-bit and 64-bit versions. Use the 32-bit (x86) dll.

You will also need the Microsoft Visual C++ 2015 Redistributable Package (vcredist_x86.exe). A link to the Microsoft installer is on the download page.

[doodal]

Quote:

Originally Posted by sanlyn

I have received this PM request from member doodal:

Please don't PM with technical questions such as this. Other users can't share this information in PM.

I don't know of an "aWarpSharp2.dll". The plugin is awarpsharp.dll. The latest version is dated 2016-06-24. The download page is here: http://avisynth.nl/index.php/AWarpSharp2. The download has 32-bit and 64-bit versions. Use the 32-bit (x86) dll.

You will also need the Microsoft Visual C++ 2015 Redistributable Package (vcredist_x86.exe). A link to the Microsoft installer is on the download page.

Installing the Microsoft Visual C++ 2015 Redistributable Package fixed it. The "2" was a typo. Thanks.

[sanlyn]

Good work. Post if any problems.

[lordsmurf]

Quote:

Originally Posted by sanlyn

IPlease don't PM with technical questions such as this. Other users can't share this information in PM.

Correct.

Allowable = "I'm having trouble, and want your input. Could you look at my post?"
And then answer in the post.

... especially on Avisynth topic! Even I learn a lot from our discussions.

[doodal]

The script works great. The only thing is cropping 10 off the bottom isn't enough to cut off the fuzziness, 16 is. Would it be ok to crop off more and make a bigger border at the bottom or is that cropping off too much picture?

[sanlyn]

Wow, how big is the frame? I haven't seen 16 pixels of head switching noise in a video in a long time. But if you say it's 16 pixels, so be it.

The original crop code was this:

Crop(12,2,-8,-10).AddBorders(10,6,10,6)

To crop off 16 bottom pixels and center the image vertically, you need this:

Crop(12,2,-8,-16).AddBorders(10,8,10,10)

The bottom border will be slightly thicker by 2 pixels, but you can't add odd-numbered pixels in YUV. I've encountered slightly unequal borders a great many times. No one notices during play. Don't forget the crop also takes 2 pixels off the top, which is an uneven border that "twitters" if it isn't evened out to 2 pixels.

[edwin_r]

Just wanted to thank you all (so much) for these tips and explanations for newbie guys.
i'm playing with HDRagc, and want more light only in darker scenes, so set coef_gain=2.0 and min_gain=1.5. Is that too much? 'cause you only set it to 1.3, and didn't specify min_gain. My video (8mm) is similar to the example, by night and artificial light.
You use HDRagc before deinterlacing (QTGMC), so SeparateFilelds() and Weave() applied. But i want to know if that filter yields worse if you only use it directly after QTGMC, (and forget about SeparateFields). i wont use FixVHSOversharp.
Thanks if you are reading.