Panasonic VCR Settings

[lollo]

Except the upscaling operation, there is no need to resize anything!

[Sharc]

TG2 and LSFMod slow down processing a lot ...

Try MCDegrainSharp() as a fast alternative to TG2 + LSFMod and leave QTGMC as suggested. Find out what suits you best.

[Traderbam]

TG2 and LSFMod slow down processing a lot ...

Try MCDegrainSharp() as a fast alternative to TG2 + LSFMod and leave QTGMC as suggested.

Nice tip, thanks. Just to clarify, do you mean QTGMC as per lollo's customised config, or one of the standard presets?

[Sharc]

TG2 and LSFMod slow down processing a lot ...

Try MCDegrainSharp() as a fast alternative to TG2 + LSFMod and leave QTGMC as suggested.

Nice tip, thanks. Just to clarify, do you mean QTGMC as per lollo's customised config, or one of the standard presets?

QTGMC as per lollo's customised config.

[lollo]

Yes, Sharc proposed a valid alternative. MCDegrainSharp() is a faster and solid solution, including denoise and sharpening. It is built on basic MVTools Mdegrain function. Attached is the version I sometime use, there are others around...

function McDegrainSharp(clip c, int "frames", float "bblur", float "csharp", bool "bsrch")
{ # Based on MCDegrain By Didee, http://forum.doom9.org/showthread.php?t=161594
  # Also based on DiDee observations in this thread: http://forum.doom9.org/showthread.php?t=161580 
  # "Denoise with MDegrainX, do slight sharpening where motionmatch is good, do slight blurring where motionmatch is bad"
  # In areas where MAnalyse cannot find good matches, the blur() will be dominant.
  # In areas where good matches are found, the sharpen()'ed pixels will overweight the blur()'ed pixels
  # when the pixel averaging is performed. 
	frames = default(frames, 2)
	bblur  = default(bblur, 0.6)
	csharp = default(csharp, 0.6)
	bsrch  = default(bsrch, true)
	bs = (c.width>960) ? 16 : 8
	
	c2 = c.blur(bblur)
	super = bsrch ? c2.MSuper(pel=2, sharp=1) : c.MSuper(pel=2, sharp=1)
	super_rend 	= c.sharpen(csharp).MSuper(pel=2, sharp=1,levels=1)
	backward_vec3 = MAnalyse(super, isb = true, delta = 3, blksize=bs, overlap=bs/2)
	backward_vec2 = MAnalyse(super, isb = true, delta = 2, blksize=bs, overlap=bs/2)
	backward_vec1 = MAnalyse(super, isb = true, delta = 1, blksize=bs, overlap=bs/2)
	forward_vec1 = MAnalyse(super, isb = false, delta = 1, blksize=bs, overlap=bs/2)
	forward_vec2 = MAnalyse(super, isb = false, delta = 2, blksize=bs, overlap=bs/2)
	forward_vec3 = MAnalyse(super, isb = false, delta = 3, blksize=bs, overlap=bs/2)	
	(frames<=0) ? c :\
	(frames==1) ? c2.MDegrain1(super_rend, backward_vec1,forward_vec1,thSAD=400) :\
	(frames==2) ? c2.MDegrain2(super_rend, backward_vec1,forward_vec1,backward_vec2,forward_vec2,thSAD=400) :\
		      c2.MDegrain3(super_rend, backward_vec1,forward_vec1,backward_vec2,forward_vec2,backward_vec3,forward_vec3,thSAD=400)
	return(last)
}

[Traderbam]

Yes, Sharc proposed a valid alternative. MCDegrainSharp() is a faster and solid solution, including denoise and sharpening. It is built on basic MVTools Mdegrain function. Attached is the version I sometime use, there are others around...

function McDegrainSharp(clip c, int "frames", float "bblur", float "csharp", bool "bsrch")
{ # Based on MCDegrain By Didee, http://forum.doom9.org/showthread.php?t=161594
  # Also based on DiDee observations in this thread: http://forum.doom9.org/showthread.php?t=161580 
  # "Denoise with MDegrainX, do slight sharpening where motionmatch is good, do slight blurring where motionmatch is bad"
  # In areas where MAnalyse cannot find good matches, the blur() will be dominant.
  # In areas where good matches are found, the sharpen()'ed pixels will overweight the blur()'ed pixels
  # when the pixel averaging is performed. 
	frames = default(frames, 2)
	bblur  = default(bblur, 0.6)
	csharp = default(csharp, 0.6)
	bsrch  = default(bsrch, true)
	bs = (c.width>960) ? 16 : 8
	
	c2 = c.blur(bblur)
	super = bsrch ? c2.MSuper(pel=2, sharp=1) : c.MSuper(pel=2, sharp=1)
	super_rend 	= c.sharpen(csharp).MSuper(pel=2, sharp=1,levels=1)
	backward_vec3 = MAnalyse(super, isb = true, delta = 3, blksize=bs, overlap=bs/2)
	backward_vec2 = MAnalyse(super, isb = true, delta = 2, blksize=bs, overlap=bs/2)
	backward_vec1 = MAnalyse(super, isb = true, delta = 1, blksize=bs, overlap=bs/2)
	forward_vec1 = MAnalyse(super, isb = false, delta = 1, blksize=bs, overlap=bs/2)
	forward_vec2 = MAnalyse(super, isb = false, delta = 2, blksize=bs, overlap=bs/2)
	forward_vec3 = MAnalyse(super, isb = false, delta = 3, blksize=bs, overlap=bs/2)	
	(frames<=0) ? c :\
	(frames==1) ? c2.MDegrain1(super_rend, backward_vec1,forward_vec1,thSAD=400) :\
	(frames==2) ? c2.MDegrain2(super_rend, backward_vec1,forward_vec1,backward_vec2,forward_vec2,thSAD=400) :\
		      c2.MDegrain3(super_rend, backward_vec1,forward_vec1,backward_vec2,forward_vec2,backward_vec3,forward_vec3,thSAD=400)
	return(last)
}

Perfect, I'll investigate how to add that to Staxrip as a function. Are the defaults ok to use with your QTGMC config?

[lollo]

Perfect, I'll investigate how to add that to Staxrip as a function.

You do not need any GUI, just write an AviSynth script and load in VirtualDub or AvsPMod.

Are the defaults ok to use with your QTGMC config?

Start from there and experiment changing csharp parameter (I wouldn't touch the others, YMMV)

[Traderbam]

Perfect, I'll investigate how to add that to Staxrip as a function.

You do not need any GUI, just write an AviSynth script and load in VirtualDub or AvsPMod.

Are the defaults ok to use with your QTGMC config?

Start from there and experiment changing csharp parameter (I wouldn't touch the others, YMMV)

I must try to set up AvsPMod and run things via script - Staxrip is very handy though.

Final question for now, in my current workflow I did TG2 denoise, the the resize to 1440 with the sharpen after. If MCDegrainSharp is a denoise and sharpen in combination, should I be adding another resharpen after upsizing?

[lollo]

If MCDegrainSharp is a denoise and sharpen in combination

Yes. Even because we use it to reduce processing time

should I be adding another resharpen after upsizing?

In general, you only sharpen once, before or after resizing (exception, when you upscale in more steps for anime content and sharpen in between).

On my material, sharpening before upscale sharpen more than sharpening after upscale, so I generally use first to reduce the sharpen strength of the filter trying to reduce halos enhacement.

[Traderbam]

Here's some samples which I'm keen to get thoughts on. Capturing same section on both a Philips VR1000 (DISTINCT mode which is EDIT on JVCs) and Panasonic 860 with 3D DNR off.

What is the preferred option to go with here, given the proposed scripts that we have been discussing? The 860 seems to have brighter hightlights (not sure if this is just when DNR off) but I had opted for the JVC versions as this seemed to look better with just a simple QTGMC deinterlace with standard presets.

For the untrained eye, what am I supposed to look for here? Pana has clearly more noise, but is this also classed as more detail in this case?

[Sharc]

I had a quick look only but I would probably shift the dark levels down by about 16 steps, something like

levels(32,1.0,235,16,235,coring=false)

[Alwyn]

720 x 576 is my starting point which isn't 4:3 - do I need to resize to 720x540 first? I'm in PAL land.

I don't bother resizing first. I just crop from the 720x576. I suppose, to be accurate, you'd resize to 768x576 then crop from there but given you're only cropping a small number of pixels, it won't make much difference.

Once you have your AVI capture, "NTSC" and "PAL" are not really applicable provided you apply the 4:3 display ratio or resize to a 4:3 frame size.

[lollo]

Here's some samples which I'm keen to get thoughts on. Capturing same section on both a Philips VR1000 (DISTINCT mode which is EDIT on JVCs) and Panasonic 860 with 3D DNR off.

I definitely prefer the Panasonic, which provides a better image and attenuate the white band defect on the left bottom of the frame:

https://slow.pics/s/GpEjb3l8

[Traderbam]

Here's some samples which I'm keen to get thoughts on. Capturing same section on both a Philips VR1000 (DISTINCT mode which is EDIT on JVCs) and Panasonic 860 with 3D DNR off.

I definitely prefer the Panasonic, which provides a better image and attenuate the white band defect on the left bottom of the frame:

https://slow.pics/s/GpEjb3l8

My thoughts were that Pana would be a better source for doing the slow deinterlace with better TG2 denoise given the clearer noise/detail to begin with. Do the highlights look overblown in your perspective?

The Philips in AUTO mode rather than DISTINCT (EDIT) was probably doing more processing so I didn't consider that as an option, even if it is defeating the purpose of a good JVC-like clone in the first place.

Not on these samples but on the tapes there is a some checkerboarding effect which the Philips seemed to hide better than the Pana, but for a better source overall I'm happy to stick with it. Pana is cleaner with NR on but probably limits scope for doing a decent denoise (if I am understanding correctly)

[Sharc]

The panny looks clearly sharper and perhaps more detailed, but the (contrast and noise enhancing) sharpening method is doubtful: it produces rather strong halos and does some kind of warping. Not sure which one is the better base for post processing .....

[lollo]

My thoughts were that Pana would be a better source for doing the slow deinterlace with better TG2 denoise given the clearer noise/detail to begin with.

Is a better image in general. With a better starting point you can reduce the "strength" of the filtering, that is always a little bit disruptive.

Do the highlights look overblown in your perspective?

Yes, a little bit:


Philips image is darker. It is always a question of personal preference.

The Philips in AUTO mode rather than DISTINCT (EDIT) was probably doing more processing so I didn't consider that as an option, even if it is defeating the purpose of a good JVC-like clone in the first place.

it is not defeating the purpose of a good JVC. Auto mode is not "intelligent" and sometimes makes disasters (there was a related thread few days/weeks ago). The good looking of the JVC is also present in edit mode, which is always preferable, especially if you plan a restoration.

Not on these samples but on the tapes there is a some checkerboarding effect which the Philips seemed to hide better than the Pana, but for a better source overall I'm happy to stick with it. Pana is cleaner with NR on but probably limits scope for doing a decent denoise (if I am understanding correctly)

Denoising is always more performant in post processing. You have 2 nice VCRs so you can choose the best device for each tape, the results can be different.

[Traderbam]

The panny is clearly sharper and more detailed, but the (contrast and noise enhancing) sharpening method is doubtful: it produces rather strong halos and does some kind of warping. Not sure which one is the better base for post processing .....

For the beginner, can you help me identify examples of the halo in the samples - to help me learn what to look for? The Pana has an AUTO, SOFT, Sharp setting in the menu but the 3D DNR button on the front only seemed to make a difference in AUTO mode. Unless I try again with SOFT and DNR off if you think there is still oversharpening.

This could just be something that is on the source tape. I believe these were recorded off a VHS-C camcorder.

[Traderbam]

My thoughts were that Pana would be a better source for doing the slow deinterlace with better TG2 denoise given the clearer noise/detail to begin with.

Is a better image in general. With a better starting point you can reduce the "strength" of the filtering, that is always a little bit disruptive.

Do the highlights look overblown in your perspective?

Yes, a little bit:

[Attachment 84919 - Click to enlarge]

Philips image is darker. It is always a question of personal preference.

The Philips in AUTO mode rather than DISTINCT (EDIT) was probably doing more processing so I didn't consider that as an option, even if it is defeating the purpose of a good JVC-like clone in the first place.

it is not defeating the purpose of a good JVC. Auto mode is not "intelligent" and sometimes makes disasters (there was a related thread few days/weeks ago). The good looking of the JVC is also present in edit mode, which is always preferable, especially if you plan a restoration.

Not on these samples but on the tapes there is a some checkerboarding effect which the Philips seemed to hide better than the Pana, but for a better source overall I'm happy to stick with it. Pana is cleaner with NR on but probably limits scope for doing a decent denoise (if I am understanding correctly)

Denoising is always more performant in post processing. You have 2 nice VCRs so you can choose the best device for each tape, the results can be different.

Good to know. The roof of the car obviously looks very strong highlight but perhaps this was just due to how it was recorded on the initial recording.

Looking at the Philips sample, would you suggest there isn't adequate detail to do the QTGMC slow, with TG2 denoise process? Is that a risky-candidate for overprocessing?

[Sharc]

For the beginner, can you help me identify examples of the halo in the samples ....

Example:

[Attachment 84920 - Click to enlarge]

To me such halos and other artifacts are more disrupting in the main object of the video than some missing sharpness of the grass and road gravel. Personal preference always matters though.
(At the end one may also check how it looks on a large TV screen as TVs add their own magic).

[Sharc]

Do the highlights look overblown in your perspective? ....

It's actually an out-of-gamut issue. The YUV histogram fits perfectly within its boundaries, but for playback the YUV gets converted to RGB and causes some clipping, indicated as cyan pixels in the right picture. It's very much the same for the Panny and the Philips btw., and it's worse (but not dramatic) for the bright sky at the beginning of the clip. Nothing unusual though. It can be on the tape (probably in this case so you can't do much about it), it can be the capturing process - we don't know for sure.


[Attachment 84922 - Click to enlarge]

[Traderbam]

Thanks Sharc, that makes sense. With checkerboarding in some other areas I think the recording process from camcorder to VHS probably introduced some of these issues which cannot be helped here.

On the Panasonic 860 I have 2 test captures - one with AUTO mode (in menu options) with 3D DNR OFF, and the other on SOFT with 3D DNR OFF. I'd describe the AUTO mode as more noisy, but perhaps it is more detail, and this mode is preferable? I didn't try SHARP as I've read elsewhere that the Panasonic's are usually oversharp by default.

Appreciate any thoughts/advice.

[Sharc]

Well, you have been given hints what to look at. Regarding details, sharpness, noise, artifacts etc there is ample room for personal preference, and only you know what looks best to you.
Technically your captures tend to have scenes with crushed brights (clipped blue) leading to discolored areas with loss of details. These may however be burnt into the tapes (source) rather than being a capturing issue (?), and then there is no real "fix" for it, so live with it.
Maybe others see it differently.

[lollo]

On the Panasonic 860 I have 2 test captures - one with AUTO mode (in menu options) with 3D DNR OFF, and the other on SOFT with 3D DNR OFF. I'd describe the AUTO mode as more noisy, but perhaps it is more detail, and this mode is preferable?

I prefer AUTO mode for your first capture bacause a sharper image: https://slow.pics/s/B7K27eTo

OTHH the halos (already present) are a bit amplified:


Levels are ok:


What Sharc pointed out is that there are illegal RGB values (in red) even after a level change to 16-235 range for proper YUV->RGB color space conversion:

[Traderbam]

Thank you gents. Incredibly useful. Will save this thread for reference/learning.

[rgr]

Here's some samples which I'm keen to get thoughts on. Capturing same section on both a Philips VR1000 (DISTINCT mode which is EDIT on JVCs) and Panasonic 860 with 3D DNR off.

What is the preferred option to go with here, given the proposed scripts that we have been discussing?

I haven't watched the clips, but I always prefer Panasonic videos to JVC (Philips). Panasonic gives a brighter image, theoretically the JVC image has more contrast, but it looks unnatural. Especially the faces, which are darker, have too much saturation. Yes, all this can be corrected later, but why if you don't have to do it with the Panasonic capture?

[Sharc]

What Sharc pointed out is that there are illegal RGB values (in red) even after a level change to 16-235 range for proper YUV->RGB color space conversion

Just to be clear on terminology (and what I used for out-of-gamut aka "illegal" RGB evaluation):
ITU-R BT-601-7 limited range YUV -> limited range RGB ("studio") equations.
Source is on the left, highlighted "illegal" RGB pixels in cyan on the right.


[Attachment 84957 - Click to enlarge]

Script (v2.0) and discussion here and above:
https://forum.doom9.org/showthread.php?p=2012882#post2012882

[lollo]

Just to be clear on terminology

Yes, we are aligned on purposes and definitions. I used jagabo's functions to highligth illegal RGB values in red https://forum.videohelp.com/threads/399404-Level()-And-Why-Does-It-Also-Modify-Chroma/...e2#post2600587

Illegal RGB values in YUV captures are common, not much you can do, except may be play a little bit with procamp input levels.

[Sharc]

Illegal RGB values in YUV captures are common, not much you can do, except may be play a little bit with procamp input levels.

Long time ago user trevlac@doom9 developped a Vdub32 filter which "legalized" out-of-gamut pixels by performing a local desaturation or similar of "hot pixels". Don't know how reliable it was. The filter has never been ported to 64bit as far as I know, and importing it in Avisynth didn't really work here.
Studios have their pro equipment to legalize their stuff, so one might find this problem less with TV captures and commercial tapes than with homemade videos.

On a sidenote I found the checkerbord type of noise of the panny captures a bit strange. Wondering whether @Traderbam used a poorly screened S-video cable causing luma<->chroma crosstalk .....

[lollo]

Long time ago user trevlac@doom9 developped a Vdub32 filter which "legalized" out-of-gamut pixels by performing a local desaturation or similar of "hot pixels".

Thanks for the info!

Studios have their pro equipment to legalize their stuff, so one might find this problem less with TV captures and commercial tapes than with homemade videos.

I confirm that with my own captures.

On a sidenote I found the checkerbord type of noise of the panny captures a bit strange. Wondering whether @Traderbam used a poorly screened S_video cable....

He can experiment with a quality S-Video cable and report back

[Traderbam]

[QUOTE=lollo;2764390]

On a sidenote I found the checkerbord type of noise of the panny captures a bit strange. Wondering whether @Traderbam used a poorly screened S_video cable....

He can experiment with a quality S-Video cable and report back

I'll look at that when I recapture. From memory it was present on captures from both VCRs, both using different cables. I'm guessing it is baked into tape but will retest.