Ghosting / fading signal in VHS captures

Malcolm · 23rd March 2008, 20:19

Hi,
while processing some old vhs captures i have encountered some artifacts which could be described as ghosting, but not like shifted copies of the signal but instead like a fading horizontal shadow of strong (black or white) amplitudes in the signal. I think the following pictures describe it better than my words:

I think these 'artifacts' are a typical result of the analogue signal processing in the vhs recorder or a form of degradation caused by the aging of vhs tapes.

Any idea of how these shadows can be removed?
Maybe it's a question of how the signal can be restored, which means it's best done in frequency space?

R3Z · 25th March 2008, 05:43

Have you tried temporal filtering ? ie Motion Compensated Median Bluring (MedianBlurT is a good one)?

Malcolm · 25th March 2008, 10:58

@R3Z
i'm not sure if i understand you correctly. Maybe we're talking about 2 different things: These shadows are no 'motion trails' or something like that. The areas where they occur are completely static. It's just that dark elements in the picture have these (fixed) 'shadows' to the right.
So i don't know how temporal filtering or MC Median blurring could help. Maybe you'd like to explain.

yup · 25th March 2008, 12:22

Malcolm!
May be problem in video cable which connecting VCR and capture card? During fast transition and edge if You use not very good cable You get ringing (cable not have enough bandpass).
yup.

2Bdecided · 25th March 2008, 12:48

I doubt it's the cable. I have tapes that look like that (Pingu mostly, so I'm not normally the one watching them!).

Cheers,
David.

yup · 25th March 2008, 13:36

2Bdecided!
Then this restriction bandpass for VHS tape not cable and need try deringing filter or dehalo.
yup.

Malcolm · 25th March 2008, 13:38

Quote:

Originally Posted by yup

Malcolm!
May be problem in video cable which connecting VCR and capture card?

I tried both my Pioneer DVD/HD-Recorder as well as a capture card connected to the VCR. the DVD/HD-Recorder actually did a better job than the capture card. I have used a very good cable to connect them to the VCR. So i don't think i can improve this further. But thanks for the hint.

yup · 25th March 2008, 14:16

Malcolm!
Try http://avisynth.org/vcmohan/FFTQuiver/F1Quiver.htm
In low pass with high value cut frequency.
yup.

R3Z · 26th March 2008, 07:47

Quote:

Originally Posted by Malcolm

@R3Z
i'm not sure if i understand you correctly. Maybe we're talking about 2 different things: These shadows are no 'motion trails' or something like that. The areas where they occur are completely static. It's just that dark elements in the picture have these (fixed) 'shadows' to the right.
So i don't know how temporal filtering or MC Median blurring could help. Maybe you'd like to explain.

I wasnt sure if they were temporal or static, but you have just stated they are static so temporal filtering isnt going to help.

Malcolm · 26th March 2008, 16:59

Quote:

Originally Posted by yup

Try http://avisynth.org/vcmohan/FFTQuiver/F1Quiver.htm

Hi yup!
I played around a bit with F1Quiver and managed to seriously screw up my video!

Seriously, after processing by F1Quiver clean parts without 'shadows' looked a lot like the bad examples i gave in my first post!

So i'm bit confused how to set up F1Quiver to deliver usable results.

yup · 26th March 2008, 17:56

Malcolm!
Upload short unprocessed clip.
yup.

Malcolm · 26th March 2008, 19:58

here are the 2 examples from above as video clips (o.k they are a bit strange, i know

):
ghosting1_cut.m2v (1.2 mb, unprocessed)
ghosting1_processed_huffy.avi (12 mb, preprocessed)
ghosting2_cut.m2v (5.6 mb, unprocessed)

Cheers,
Malcolm

yup · 27th March 2008, 08:32

Malcolm!
If I right understandt we have low pass filtering before writing to VHS in luma and especialy in chroma space (chroma place in high frequency range and more filtered).
Low paas filter frequency responce:

pass
----------
********\
******** \
*********\ cut
And we need conctruct filter with responce:
****************increase level at high frequency
pass********/------------
********** /
**********/
saturation at high frequency need for nonsense to noise.
This filter can constructed in CCE during encoding. You need find only cut frequency.
yup.

yup · 27th March 2008, 13:56

Malcolm!
Try:

Code:

ChromaShift(C=-2)

http://www.geocities.com/siwalters_uk/fnews.html
See little better than source.
yup.

Malcolm · 27th March 2008, 20:41

@yup
Thanks for your effort. When using Chromashift(C=-2) i get the following output:

So this filter does only change the chroma channels. But the shadowing is a problem in the luma channel. What do you try to achieve with this filter?

yup · 28th March 2008, 08:59

@Malcolm!
I think that problem coupled shifting between luma and chroma planes and at edges chroma have small delay. I use chromashift on second clip and see small improvement. Also I do not know filter for increase high frequency (only in CCE), F1Quiver can not made it.
yup.

2Bdecided · 28th March 2008, 11:22

Here's an idea. The problem appears to be like the application of a very strong horizontal low pass filter, and then mixing the output of this filter very slightly into the original video signal.

Maybe you could apply a very strong horizontal low pass filter, and then subtract the result (scaled down a lot) from the original video signal. If you can get the low pass filter and the scaling correct (i.e. to match the original problem) you should see a big improvement.

I'll leave it to the AVIsynth scrop gurus to figure out how to actually implement it

Cheers,
David.

Malcolm · 28th March 2008, 14:57

Quote:

Originally Posted by 2Bdecided

Here's an idea. The problem appears to be like the application of a very strong horizontal low pass filter, and then mixing the output of this filter very slightly into the original video signal.

Hi David,
that's what i have in mind too. But i see one problem: A low pass filtered clip would have the signal 'smeared' horizontally in both directions. Since these 'shadows' are only visible on the right side of strong black image elements, how could it be avoided to corrupt the image left of them?
Until now i wasn't able to create a filter/scripting combination that accomplishes this.

I have another unscientific view to the problem: If you think of these shadows as an analogy to harmonic waves or echoes or 'overtones' to the main signal in frequency space. Then maybe there's an algorithm that simply can 'deducted' them out of the signal. Maybe in frequency space this is an easy equation.

2Bdecided · 28th March 2008, 15:59

You're assuming a symmetric impulse response low pass filter. There's no reason why you would have to use that - you could use something that matches the original analogue effect. For each pixel, create a weighted sum of the pixels to the left. It's just a different impulse response, but a valid filter.

I have no idea how to do this in AVIsynth. I may have heard talk of an arbitrary filter function, but I may have been dreaming!

Cheers,
David.

23rd March 2008, 20:19	#1 \| Link
Malcolm Registered User Join Date: Sep 2002 Location: Germany Posts: 352	Ghosting / fading signal in VHS captures Hi, while processing some old vhs captures i have encountered some artifacts which could be described as ghosting, but not like shifted copies of the signal but instead like a fading horizontal shadow of strong (black or white) amplitudes in the signal. I think the following pictures describe it better than my words: I think these 'artifacts' are a typical result of the analogue signal processing in the vhs recorder or a form of degradation caused by the aging of vhs tapes. Any idea of how these shadows can be removed? Maybe it's a question of how the signal can be restored, which means it's best done in frequency space?

26th March 2008, 19:58	#12 \| Link
Malcolm Registered User Join Date: Sep 2002 Location: Germany Posts: 352	here are the 2 examples from above as video clips (o.k they are a bit strange, i know ): ghosting1_cut.m2v (1.2 mb, unprocessed) ghosting1_processed_huffy.avi (12 mb, preprocessed) ghosting2_cut.m2v (5.6 mb, unprocessed) Cheers, Malcolm Last edited by Malcolm; 26th March 2008 at 20:15.

27th March 2008, 13:56	#14 \| Link
yup Registered User Join Date: Feb 2003 Location: Russia, Moscow Posts: 854	Malcolm! Try: Code: ChromaShift(C=-2) http://www.geocities.com/siwalters_uk/fnews.html See little better than source. yup.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

25th March 2008, 05:43	#2 \| Link
R3Z Silver Über Alles Join Date: Aug 2006 Location: Australia Posts: 269	Have you tried temporal filtering ? ie Motion Compensated Median Bluring (MedianBlurT is a good one)?

25th March 2008, 10:58	#3 \| Link
Malcolm Registered User Join Date: Sep 2002 Location: Germany Posts: 352	@R3Z i'm not sure if i understand you correctly. Maybe we're talking about 2 different things: These shadows are no 'motion trails' or something like that. The areas where they occur are completely static. It's just that dark elements in the picture have these (fixed) 'shadows' to the right. So i don't know how temporal filtering or MC Median blurring could help. Maybe you'd like to explain.

25th March 2008, 12:22	#4 \| Link
yup Registered User Join Date: Feb 2003 Location: Russia, Moscow Posts: 854	Malcolm! May be problem in video cable which connecting VCR and capture card? During fast transition and edge if You use not very good cable You get ringing (cable not have enough bandpass). yup.

25th March 2008, 12:48	#5 \| Link
2Bdecided Registered User Join Date: Dec 2002 Location: UK Posts: 1,673	I doubt it's the cable. I have tapes that look like that (Pingu mostly, so I'm not normally the one watching them!). Cheers, David.

25th March 2008, 13:36	#6 \| Link
yup Registered User Join Date: Feb 2003 Location: Russia, Moscow Posts: 854	2Bdecided! Then this restriction bandpass for VHS tape not cable and need try deringing filter or dehalo. yup.

25th March 2008, 14:16	#8 \| Link
yup Registered User Join Date: Feb 2003 Location: Russia, Moscow Posts: 854	Malcolm! Try http://avisynth.org/vcmohan/FFTQuiver/F1Quiver.htm In low pass with high value cut frequency. yup.

26th March 2008, 17:56	#11 \| Link
yup Registered User Join Date: Feb 2003 Location: Russia, Moscow Posts: 854	Malcolm! Upload short unprocessed clip. yup.

27th March 2008, 08:32	#13 \| Link
yup Registered User Join Date: Feb 2003 Location: Russia, Moscow Posts: 854	Malcolm! If I right understandt we have low pass filtering before writing to VHS in luma and especialy in chroma space (chroma place in high frequency range and more filtered). Low paas filter frequency responce: pass ---------- ******\ **** \ *****\ cut And we need conctruct filter with responce: ************increase level at high frequency pass****/------------ ****** / ********/ saturation at high frequency need for nonsense to noise. This filter can constructed in CCE during encoding. You need find only cut frequency. yup.

27th March 2008, 20:41	#15 \| Link
Malcolm Registered User Join Date: Sep 2002 Location: Germany Posts: 352	@yup Thanks for your effort. When using Chromashift(C=-2) i get the following output: So this filter does only change the chroma channels. But the shadowing is a problem in the luma channel. What do you try to achieve with this filter?

28th March 2008, 08:59	#16 \| Link
yup Registered User Join Date: Feb 2003 Location: Russia, Moscow Posts: 854	@Malcolm! I think that problem coupled shifting between luma and chroma planes and at edges chroma have small delay. I use chromashift on second clip and see small improvement. Also I do not know filter for increase high frequency (only in CCE), F1Quiver can not made it. yup.

28th March 2008, 11:22	#17 \| Link
2Bdecided Registered User Join Date: Dec 2002 Location: UK Posts: 1,673	Here's an idea. The problem appears to be like the application of a very strong horizontal low pass filter, and then mixing the output of this filter very slightly into the original video signal. Maybe you could apply a very strong horizontal low pass filter, and then subtract the result (scaled down a lot) from the original video signal. If you can get the low pass filter and the scaling correct (i.e. to match the original problem) you should see a big improvement. I'll leave it to the AVIsynth scrop gurus to figure out how to actually implement it Cheers, David.

28th March 2008, 15:59	#19 \| Link
2Bdecided Registered User Join Date: Dec 2002 Location: UK Posts: 1,673	You're assuming a symmetric impulse response low pass filter. There's no reason why you would have to use that - you could use something that matches the original analogue effect. For each pixel, create a weighted sum of the pixels to the left. It's just a different impulse response, but a valid filter. I have no idea how to do this in AVIsynth. I may have heard talk of an arbitrary filter function, but I may have been dreaming! Cheers, David.