Blocky edges around VHS Captures - see screenshots

[themaster1]

From what i understand you have used 3rd party softwares, what happen if you use the provided app? Let us know

[intracube]

sanlyn; when you decode the .mpg with VirtualDub or similar, (without any further processing) which image on post #27 does it look closest to?

Is this some weird decoding problem with Windows?

[jagabo]

sanlyn; when you decode the .mpg with VirtualDub or similar, (without any further processing) which image on post #27 does it look closest to?

The one on the right is closest to what VirtualDub shows. Mplayer is exacerbating the problems (along with other post processing) on the left.

Is this some weird decoding problem with Windows?

The MPG video was obviously upscaled, and poorly so, from some lower resolution (probably 352x576 or 360x576.) before being encoded. But the post processing done on your left image could be mplayer or the graphics card.

[intracube]

sanlyn; when you decode the .mpg with VirtualDub or similar, (without any further processing) which image on post #27 does it look closest to?

The one on the right is closest to what VirtualDub shows. Mplayer is exacerbating the problems (along with other post processing) on the left.

No, the image on the right is from mplayer. The image on the left is li21's VirtualDub output from post #6.
EDIT: ^this is post #27, btw. Post #30 shows mplayer's output on the left with the slight blockies, and then filtered on the right.

So the question remains - why is li21 seeing worse quality than us?

Is this some weird decoding problem with Windows?

The MPG video was obviously upscaled, and poorly so, from some lower resolution (probably 352x576 or 360x576.) before being encoded.

That would be a possibility if we were all seeing the same as li21 (post #6). But the video played back on my system (and yours, If I understand correctly) doesn't have this bad interpolation/upscaling. With a minimal amount of filtering, all the blockiness can be removed (see post #30).

[intracube]

Or am I misunderstanding the deal here?

[jagabo]

li21's image in post 6 must have had some other processing done to it. Regarding the upscaling: when you enlarge the sample you can see a clear pattern where horizontal pairs of pixels have similar luma. Keep in mind that VHS has a horizontal resolution of roughly 360 lines across the width of the frame. So upscaling a 360x576 cap to 720x576 isn't losing much of the VHS picture information. It's just that the way the upscale was done was poor. And yes, a light horizontal blur alleviates that bad upscaling.

[sanlyn]

I go along with jagabo. The rescaling is faulty, as is the encoder.

sanlyn; when you decode the .mpg with VirtualDub or similar, (without any further processing) which image on post #27 does it look closest to?

The posted mpg sample I worked with looks like the image on the right in post #27. The version in post #30 that jagabo refers to has less garbage than some of the other versions (but the garbage that was there is sharpened beyond repair). The version I ended up with is shown in post #20 -- it looks the same way as that image when viewed in VLC player.

I think the main point is being overlooked. You're saying that one image looks "better" than another, which is true but for various reasons thru different means of capturing, filtering, encoding, media player used, etc. The "better" video sample I worked with still had real problems, many of which could have been avoided with a better capture setup. It was a mistake to capture noisy, damaged tape directly to MPEG in the first place; it takes a lot of destructive filtering to clean encoded artifacts. IMO the tape player appears to be decent (few line timing wiggles or distortions). So I'd bet the major obstacles are the capture device and the encoder. How it looks in various players can be debated forever.

The way I processed the sample video can be debated as well. But it had nothing to do with the way it looks in media players, none of which are that great anyway. I used this:

MPEG2Source(vidpath+"Captureh.d2v")
maa()
SeparateFields()
top=SelectEven().RemoveSpots().deVCR(30)
bot=SelectOdd().RemoveSpots().deVCR(30)
tb=Interleave(top,bot).Weave()

tb
MCTemporalDenoise(settings="low",AA=true,DeBlock=true,sigma=6,edgeclean=true,interlaced=true)
AssumeTFF().QTGMC(preset="faster")
MergeChroma(fft3dfilter(sigma=.1, sigma2=.3, sigma3=.5, sigma4=20, plane=3))
mergechroma(aWarpSharp(depth=20.0, thresh=0.75, blurlevel=2, cm=1))
FineSharp(sstr=0.2)
LSFMod(strength=75,smode=3,secure=true,edgemode=2)
AddGrainC(var=1.5,uvar=1.5)

AssumeTFF().SeparateFields().SelectEvery(4,0,3).Weave()
ReplaceFramesMC(78,2)
ConvertToRGB32(matrix="Rec601",interlaced=true)
Crop(0,0,-2,-10).AddBorders(2,4,0,6)
return last

In VirtualDub I tweaked gamma and cleaned a little more of the "simmering" tape noise. The video would look cleaner, more detailed and more 3-dimensional if it had been captured to lossless AVI with a better capture device.

[themaster1]

Try dscaler to cap your tape it has its own drivers (again not sure it work great on win7)

[intracube]

I think the main point is being overlooked. You're saying that one image looks "better" than another, which is true but for various reasons thru different means of capturing, filtering, encoding, media player used, etc. The "better" video sample I worked with still had real problems, many of which could have been avoided with a better capture setup.

Yep, post #6 confused me - I thought li21 was showing the output of VirtualDub, unscaled and unprocessed and that he/she was getting that quality "no matter what software".

I agree on your other points.

[li21]

This forum is really amazing..Really amazes how much stuff you guys know and how you all willing to work on samples etc.
thanks!

Sanlyn, your mpg sample to me looks really clean.. i'd be happy with results like that! I might have to give avisyntch a go..

Now.. you guys are probably going to face palm.. but fundamentally what you guys are saying is correct.. the capture was bad to begin with.. whether it be the hardware or the set up..

This is what I noticed .. the last few captures.. you may have read about me complaining why the sound is squelchy and sounds like its 22khz / 96kpbs . quality even though the capture is set up at 44khz/256kbps..
I was thinking.. if the AUDIO is BAD.. despite what the capture program says it is capturing at..
then chances are it is really not captureing at the res it is saying it is... ie capturing at 320x576 not 720x576 and the upscale kills the edges and introduces artifacts .. which I am reporting in post #1..


So what did I do.. I set up a Virtual Machine on the SAME PC running Windows XP.. Install the capture drivers for the Kaisser Baas.. Installed Power Director.. Ran a capture.. and what do you know.. it looks better and sounds better.. only problem was it was a VM and dropped a heap of frames..

Next I thought .. I would go back to my main / host PC..
I uninstalled the drivers..
Instead of using the ones I downloaded.. I installed off the CD that came with it.

Did a capture this morning.. and I believe we have improvement!!

Its a bit premature to say the drivers were "locking" up the capture settings despite what the APP was telling it to capture at..

Will try again when I get home and if I still get better caps.. then problem solved for the most part..

Its not perfect however..
Its a definite improvement over the frame gram I showed you in post #6..
I think it is really doing 720x576 now..

I think applying Sanlyn's script I can get something I'll be happy with.

Themaster1 - Will try descaler too!

Other concern is.. does Windows 7 have something about it that prevents proper captures at 720x576?
You guys have mentioned issues with Windows 7?

[jagabo]

does Windows 7 have something about it that prevents proper captures at 720x576?

No. If there's a problem it's in the drivers or capture program.

[sanlyn]

I think applying Sanlyn's script I can get something I'll be happy with.

Themaster1 - Will try descaler too!

Thanks for the compliment, but be mindful...the script was set up for a poorly rescaled and encoded source. To clean up a better capture made to lossless AVI instead of lossy-encoded format, that script would be overkill (except for some really crappy tape). And it ran a very sloooowww 1-fps on a 3.5GHz quad4 system. A cleaner capture would have filtered much more quickly and required fewer demolition tools.

Another tip: don't capture to compressed audio. When folks capture to lossless AVI with huffyuv or Lagarith they capture audio as uncompressed PCM -- because processing and re-encoding lossy audio entails the same quality loss as re-encoding compressed video.

I do hear some occasional sibilance in that clip. But that could have been the broadcast.

[li21]

PART 2

So.. having watched some of my captures.. I'm still realizing that this blocky edge problem is still apparent..
Not as bad a before.. but some scenes are still unwatchable as faces and detail get lost..

So I thought I'd dump the Kaisser Baas and went and got a Compro C200 Video Mate USB device...

Immediately the blocky edge issue was gone!!!
I was getting almost like for like VHS quality which I was happy about...

BUT.. and there always is a BUT..
there seems to be more video judder and tape distortion!?

On the Kaisser Baas I didnt get that!? Seems to soak up the judder and static

On the comproo C200 I get non stop judder around 1/3 down the screen .. so at times it seems some of the top of the picture is misaligned with the bottom .. almost like its a bad tape..
Any ideas why I see this on the C200 but not on the Kaisser?

Attached sample..
(pretty minor in this sample but you can still see it )

[jagabo]

Why is your video 100 fps? Every frame is repeated four times.

[li21]

Why is your video 100 fps? Every frame is repeated four times.

Hmm no idea about that.. Maybe it was something handbrake did when I tried to
Convert to something small enough to upload..

The original capture is at 25fps though

[jagabo]

Oh sorry, the frame rate issue was at my end. For some reason VirtualDub reported the MKV file's variable frame rate as 100 fps.

You'll need a line time base corrector to fix the left/right wiggle. The levels are terrible. It looks like the capture was at half D1 (352x576). It's been way over filtered, over compressed, and cropped.

[li21]

Agree but that was just to get a file small enough to show you guys the "wiggle"

Is it possible the capture device is faulty , causing that wiggle?

For a newbie like me is there any line time correction that's simple enough to use without scripts ?

[jagabo]

All I can tell you is different capture devices have different susceptibility to time base errors.

[sanlyn]

For a newbie like me is there any line time correction that's simple enough to use without scripts ?

Time base correctors are hardware devices, not software. Line and frame timimg errors can't be corrected with software, nor can they be corrected after capture or re-recording.

[jagabo]

In analog video the start of every scan line is marked with a sync pulse:

http://en.wikipedia.org/wiki/Analog_television#Structure_of_a_video_signal

In a studio quality signal the timing of those sync pulses is very accurate. But the speed of the spinning drum in a VCR is not so accurate. It is continuously being adjusted. Sometimes it runs a little too fast, sometimes a little too slow. That means the time between the sync pulses varies slightly. Most capture cards do not attempt to adjust for that variation. That means you need to fix the signal before giving it to the capture card. That is what a line TBC does. It captures each scan line, adjusts its length, and sends it out with perfect timing (in practice, the variation of the drum speed isn't linear so there's still a little skew in the picture content, but it's much cleaner than what the VCR put out). The line TBC can be a part of the VCR (S-VHS decks often include it) or an external device.

Some old DVD recorders have line TBCs (and full frame TBCs) that can be used in pass-through mode as a poor man's line TBC.

https://forum.videohelp.com/threads/319420-Who-uses-a-DVD-recorder-as-a-line-TBC-and-what-do-you-use