Noise removal

[Tafflad]

Been doing lots of VHS (PAL) digitization, and then processing for online streaming. (final format H264/mp4 progressive)

This has raised 2 question ....

First I have a number of scripts that are for denoising VHS ... 2 attached

DeNoise Interlaced VHS -- MDegrain2, Multi-threaded, Mrecalculate, optional sharpening.avs
latest johnmeyer denoiser_progressive.avs

I note one is dated 2015, the other is quoted as latest but no idea on the date?

Could somebody who understands these scripts advise which is the latest and best script to use.


Secondly -
The one entitled 'denoiser_progressive' I had been using after deinterlace with QTGMC
Should I stick to that approach .... or use it before deinterlacing.
I have tried it before and after deinterlacing and can see little difference.


Or am I better to use DeNoise Interlaced VHS and run that before deinterlacing.

[johnmeyer]

Both of those are my scripts.

Do NOT deinterlace. Deinterlacing degrades your video by throwing away 1/2 of all temporal information. Would you throw away 1/2 your video's spatial resolution by eliminating every other line? Of course not. So, don't do the equivalent thing in the time domain.

My scripts properly handle interlaced material by treating odd fields as one video and even fields as a different video and then combining them back together.

The only time I ever deinterlace is if I have to re-size the video. For this operation deinterlacing is mandatory. I also deinterlace if the final result is being uploaded to a streaming service (e.g., YouTube). For everything else, I keep it interlaced in order to not needlessly degrade the result. Interlaced video is not in any way "inferior" to progressive, a common and totally erroneous misconception. While it is true that interlaced video requires a few extra steps to properly restore, if those steps are done correctly, the result is just as good as starting with progressive, and by not throwing away all that temporal information, you end up with a video that is as smooth as the original, without deinterlacing artifacts.

The first script you listed is the latest version of my efforts to improve VHS. However it is in no way "best." Over at doom9.org they don't even let you talk about "best" and it is a good rule. The reason for the rule is that every single video is different, and a script that works wonderfully for one video may actually make a different video look terrible. (Also, not mentioning "best" avoids a lot of pointless arguments.)

You may prefer my earlier script that didn't use MRecalculate. I've copied it below. Whichever script you use, you need to play around with the settings. The block size and overalap make a big difference in the final result. You can use block size of 4, 8, or 16. Overlap must be a power of 2, and can be no larger than 1/2 the block size (e.g., for block size 16 you can have an overlap of 2, 4, or 8). Play close attention not only to any fine details lost, but also to how large flat areas are treated. These scripts can alter big flat areas, like blue sky or paved roads, and remove some of the subtle contours. Proper adjustment of the parameters can minimize these losses.

You must remove the SetMTMode calls if you don't use the multi-threaded version of AVISynth. The single-thread version will run at about 1/3 speed of the multi-threaded one.

Don't use DCT=1 unless your video has a lot of flicker. Setting DCT to anything but zero will quadruple the time it takes the script to run, and will not improve the noise reduction for VHS material.

Finally, don't bother with using the third level of estimation followed by using MDegrain3. I have never once seen any improvement with VHS material.

Earlier script

#Denoiser script for interlaced video using MDegrain2

SetMemoryMax(768)

Loadplugin("C:\Program Files\AviSynth 2.5\plugins\MVTools\mvtools2.dll")
LoadPlugin("c:\Program Files\AviSynth 2.5\plugins\CNR\Cnr2.dll")

SetMTMode(5,6)
#Modify this line to point to your video file
source=AVISource("E:\fs.avi").killaudio().AssumeBFF()
SetMTMode(2)

#Only use chroma restoration for analog source material
chroma=source.Cnr2("oxx",8,16,191,100,255,32,255,false) #VHS
#chroma=source.Cnr2("oxx",8,14,191,75,255,20,255,false) #Laserdisc 
#Set overlap in line below to 0, 2, 4, 8. Higher number=better, but slower
#For VHS, 4,0 seems to work better than 8,2. Most of difference is in shadows
#However, 8,0 is good enough and MUCH faster. 8,2 doesn't seem to make much difference on VHS.

#output=MDegrain2i2(chroma,8,0,0)
output=MDegrain2i2(chroma,8,4,0)  #Better, but slower

#stackvertical(source,output)
#stackhorizontal(source,output)
return output


#-------------------------------

function MDegrain2i2(clip source, int "blksize", int "overlap", int "dct")
{
Vshift=0 # 2 lines per bobbed-field per tape generation (PAL); original=2; copy=4 etc
Hshift=0 # determine experimentally 
overlap=default(overlap,0) # overlap value (0 to 4 for blksize=8)
dct=default(dct,0) # use dct=1 for clip with light flicker

fields=source.SeparateFields() # separate by fields

#This line gets rid of vertical chroma halo
#fields=MergeChroma(fields,crop(fields,Hshift,Vshift,0,0).addborders(0,0,Hshift,Vshift))
#This line will shift chroma down and to the right instead of up and to the left
#fields=MergeChroma(fields,Crop(AddBorders(fields,Hshift,Vshift,0,0),0,0,-Hshift,-Vshift))

super = fields.MSuper(pel=2, sharp=1)
backward_vec2 = super.MAnalyse(isb = true, delta = 2, blksize=blksize, overlap=overlap, dct=dct)
forward_vec2 = super.MAnalyse(isb = false, delta = 2, blksize=blksize, overlap=overlap, dct=dct)
backward_vec4 = super.MAnalyse(isb = true, delta = 4, blksize=blksize, overlap=overlap, dct=dct)
forward_vec4 = super.MAnalyse(isb = false, delta = 4, blksize=blksize, overlap=overlap, dct=dct)

#Increasing thSAD doesn't seem to help
 MDegrain2(fields,super, backward_vec2,forward_vec2,backward_vec4,forward_vec4,thSAD=400) 

#UnsharpMask( clip , int "strength" , int "radius" , int "threshold" ) 
#strength: strength. The default is 64.
#radius: the scope of the blurring process. The default is 3.
#threshold: threshold. Absolute value of the processing component is greater than the threshold blur. The default is 8.

#unsharpmask(60,3,0) #not sure whether to put this before or after the weave.

Weave()
}

[DB83]

But the OP already stated that:

1. The output is for on-line streaming and
2. The source is PAL. (where yt etc. does not natively support 576 lines and will resize with often tragic results)

So surely by your own words he MUST de-interlace

[johnmeyer]

But the OP already stated that:

1. The output is for on-line streaming and
2. The source is PAL. (where yt etc. does not natively support 576 lines and will resize with often tragic results)

So surely by your own words he MUST de-interlace

Thank you once again for another wonderful post, in which you manage to avoid helping the OP in any way.

Yes, I understand the output is for on-line streaming. However, you are assuming that he has no other use for the video. It is better to keep the video in its original state for as long as possible during the restoration process to keep open all your options open later on. For instance, if he wants to later create a DVD or Blu-Ray, he can still do that if he doesn't deinterlace.

Another point: online streaming sites know how to handle deinterlacing because they receive interlaced video all the time. It has been my experience that most people -- even those who otherwise "know what they're doing -- often screw up the deinterlacing step, whereas the streaming site will get it right. Having said that, I would agree that if you DO know how to properly deinterlace, you will probably get a better result if you do it yourself prior to uploading, and YouTube themselves recommend that.

Finally, why don't you actually try to help this person instead of once again making posts that have no reason other than to annoy (troll?). I'll bet you can't do it. Let's see if you can actually help this person by providing some suggestions that will help him get a better result. My bet is that you can't do it, or at least won't be able to do it without making some comment about me or my posts that you think is really clever but, of course, is not.

Oh yes, your point about YouTube not supporting PAL is wrong:

YouTube Encoding (from Google/YouTube site)

It was briefly true back in 2010, but not any longer.

[manono]

And you, johnmeyer, are being unnecessarily harsh in response to a well intended post by DB83, in my opinion. I don't know if the two of you have a 'history' or not.

Yes, I understand the output is for on-line streaming. However, you are assuming that he has no other use for the video.

And you're assuming there is another use. The questions were purely about deinterlacing for online streaming.

...online streaming sites know how to handle deinterlacing because they receive interlaced video all the time.

And why would I want YouTube or any other place do the deinterlacing for me when I can do a much better job myself? And, yes, I understand you made the same point yourself, but it looks to me that you used the comment as a jumping off point to insult DB83, whose own deinterlacing abilities aren't the issue here. Again, though, none of this has anything to do with Tafflad's question.

Oh yes, your point about YouTube not supporting PAL is wrong...

No, it wasn't. His point was about 576p not being shown on YouTube and not about allowable framerates.

Since you're always so cheerful and helpful and not prone to insulting behavior usually, I'm surprised at your response so there must be more going on here than meets the eye. Of course, since I sometimes am rude myself, I'm probably the last one who should be preaching about online etiquette.

[DB83]

Actually the OP is a friend and he considers my posts as a means to move the discussion forward and save him the time of the confusion created in your initial reply.

And why should you assume that he wants to create dvds as well? Surely the whole point of a response is to deal with the question(s) and not add more.

Now if you consider that as 'trolling' then I really am sorry for you.

And for my part I also want to 'learn' since my own attempts in uploading interlaced content to yt was not good. Don't ask me when that was.

[DB83]

@manono

Our 'expert' considers I insulted him some months ago. Since then He does not like me joining any thread he has contributed in. Even if I have only the interest of the OP in mind.

Rather sad methinks.

[johnmeyer]

And you, johnmeyer, are being unnecessarily harsh in response to a well intended post by DB83, in my opinion. I don't know if the two of you have a 'history' or not.

Yes, you are correct. You need look no further than his last response:

Our 'expert' considers I insulted him some months ago.

You may think I was reading between the lines and taking too much from how he posted above, but surely no one can miss the dismissive attitude inherent in putting the word "expert" in quotes: it means he considers me to be anything but an expert.

I got off on a bad foot with him last January when he tried to help me -- and he was initially being very helpful -- but as the thread progressed, something got lost in the translation and he ended up coming back at me with this:

To be totally frank and this time it will be my last word since you are beyond help - the excuse for not posting that capture is purile - you did not look for help in creating this topic.

If you read that thread, that statement was completely false. I was tracking down what turned out to be a bug in the Hauppauge drivers, and the best example of the problem turned out to be a clip from a Showtime "National Lampoon" documentary that contained nudity, and I did not want to post it because I didn't want to run afoul of community standards.

And, as I predicted in my post above, he has made no attempt above to actually help the OP with his denoising issue. I would hope that perhaps someone could suggest a better method to get his video onto YouTube, and provide a workflow that would do a better job than my "non-expert" scripts. I make no claim to knowing more than anyone else, although in utter frustration in that earlier discussion I did briefly attempt to defend my efforts by mentioning that I had, at one time, some association with the software industry.

To the OP: I apologize for not helping you and for being a party to taking this discussion off topic. Since DB83 is a close friend of yours, and since he really does know quite a bit, perhaps the two of you should work together on this.

[DB83]

Oh jeezuz,

I NEVER claimed to knowing. And it might help manono's understanding of your animosity to me if you actually posted a link to that topic. No need to since I already did.

And no where do I recall any mention of a 'association with the software industry' unless you translate that as running various companies. Now by your twisted interpretation of the use of quotes I guess you will now think I mean that you never had such an association.

Now let me, but why I should I know not, explain your 'expert' status here. The OP, a friend yes (close is a word again you added) linked to two scripts. You then jumped in correctly claiming the authorship of them. So that itself makes you an 'expert'.

But even an 'expert' can get it wrong. Do you still contend that the OP should upload 576i footage to yt ? That was my mistake since the end result is de-interlaced, resized footage with visible artifacts. Maybe yt is kinder to 576i footage now. Yet most uploads I guess are 480i so no vertical resizing is required.

Have you ever dealt with 576i footage ?

BTW sarcasm is the lowest form of wit.

[johnmeyer]

I NEVER claimed to knowing.

I don't know what this means.

And it might help manono's understanding of your animosity to me if you actually posted a link to that topic. No need to since I already did.

I have no problem posting a link, I just didn't get around to it. You say that posted a link? I don't see it in your message.

And no where do I recall any mention of a 'association with the software industry' unless you translate that as running various companies.

What is your point?? I don't even understand what you are trying to say. Yes, I was referring to the fact that I ran several software companies. However, I realized that when I said that back in January, it made me sound supercilious, so I was trying to be a little more discrete this time around.

Now by your twisted interpretation of the use of quotes I guess you will now think I mean that you never had such an association.

I do not understand what "twisted interpretation" is supposed to mean. Twisted ??? I guess not everyone understands every last nuance of the English language, but in case you don't know, quotes around a word, when used the way you did, indicates sarcasm. Here's one reference: Using Quotation Marks to Show Sarcasm. Scroll about 1/3 of the way down the page to see the example.

But even an 'expert' can get it wrong.

Quite true. I never claimed to be an expert (that was your word), but even if I did, I will readily admit that I get things wrong all the time.

Do you still contend that the OP should upload 576i footage to yt? That was my mistake since the end result is de-interlaced, resized footage with visible artifacts. Maybe yt is kinder to 576i footage now. Yet most uploads I guess are 480i so no vertical resizing is required.

Have you ever dealt with 576i footage ?

First of all, yes, I deal with PAL SD interlaced footage (576i) fairly frequently. However, in answer to your first question, I'll admit that I never tried uploading an interlaced PAL clip directly to YouTube. So, in order to find out if it is a problem, I just took a DV AVI clip that is directly from a PAL SD camera, and uploaded it to YouTube, as is.

Here's the result:

https://youtu.be/rZhCMMqTyfY

It looks just fine to me, almost exactly like the original. This clip is very, very noisy, so don't mistake the sensor noise for YouTube artifacts. The reason I have this video is that I wanted to help the person remove that noise without harming the details on the fish. Here's a link to a very long thread where several of us worked together to create some really interesting results: My Latest Denoising Tests

So, I stand by everything I said in the original post, including my statement that many streaming services are perfectly capable of handling interlaced video uploads. I think my brief test proves that out. Having said that, YouTube's upload instructions do recommend that you upload progressive. However, since 99% of all people uploading to YouTube don't have the slightest idea about interlaced vs. progressive, YouTube obviously has to be able to deal with interlaced video, and it appears that they can now do that for PAL as well as NTSC interlaced clips.

BTW sarcasm is the lowest form of wit.

I couldn't agree more, which is why I reacted so strongly to your sarcastic quotes.

[edit] In the end, I'm just trying to help this person, which I hope I was able to do. So, give me a break, please.

[manono]

Could somebody who understands these scripts advise which is the latest and best script to use.

For online streaming I'm a fan of the FFT3DFilter. It filters both spatially and temporally. You might or might not use it for your own encodes but for streaming where YouTube doesn't allow for enough bitrate for your videos (unless it's just talking head type videos), it makes the videos very compressible so that they look pretty good on YouTube. A simple line such as:

fft3dfilter(sigma=2.0,bt=5,bw=32,bh=32,ow=16,oh=16 )

will clean up a video nicely. Raise the Sigma value for more cleaning. It can also sharpen and dehalo. Because of the low bitrates used by these sites, trying to retain detail/sharpness is self-defeating in many cases, depending on the kind of content. The video will sometimes wind up looking worse than if you had used a stronger cleaner to begin with, one such as the FFT3DFilter. Because it's also a spatial/temporal cleaner, I like to use DFTTest sometimes. I make everything progressive before cleaning. You can even use DFTTest and FFT3DFilter within QTGMC.

[Tafflad]

Gents ... sorry to have kicked off an argument ... I have no desire to have sparked that with either of you.

Perhaps I could explain further, the questions raised are specifically in relation to creating of progressive files from SVHS PAL captures (720x576) for processing and uploading to Vimeo & Youtube for streaming.
(For non-streaming I would not deinterlace)

I also want to Deinterlace in the workflow as many of these files will have DeShaker applied .. and multiple comments on this forum advise that DeShaker is better on progressive video.

The workflow had been:

Capture to DV
Rough edit > intermediate files saved as Magic YUV (video only)
Deinterlace & resize to 960x720
Denoise
Deshake & sharpen (if required)
Final edit, render as H264/MP4 (960x720,50p) with AAC audio

My question is in 2 parts.

The first was - should I use your script before or after deinterlacing, I understand your explanation........... does that mean that the script "latest johnmeyer denoiser_progressive" is in fact wrongly named and should only be used on interlaced material ? or did you modify it specifically for progressive ?


Part 2
I will make sure in future I use your scripts 'before' I deinterlace. Assume I would be better served using the one you posted above - I won't call it best I'll try that out (before deinterlacing)


I don't claim to understand a great deal of the stuff discussed in these forums, and sometimes I ask wrong questions, sometimes not even knowing what to ask .... and I appreciate the assistance form those that know.

This post started as I have come across many versions of the John Meyer 'denoise' scripst and knowing how these evolve was trying to get to the latest and 'possibly' most appropriate. (see learnt not to use the word best)

[jagabo]

If you're going to deinterlace with QTGMC and are also looking for noise reduction you should try using its built in noise reduction options. For example:

QTGMC(preset="slow", EZDenoise=2.0, Denoiser="dfttest", DenoiseMC=true)

[johnmeyer]

I also want to Deinterlace in the workflow as many of these files will have DeShaker applied .. and multiple comments on this forum advise that DeShaker is better on progressive video.

If you Google my name along with "deshaker" you will find that I wrote an extensive user guide for Deshaker. In addition, I also wrote a set of Vegas and VirtualDub scripts to completely automate the process of using Deshaker with Sony Vegas. As a result of both of these projects, I have a very good understanding of that software, and I can tell you that it works perfectly fine with interlaced footage. I have stabilized dozens of hours of interlaced footage with it over the past ten years.

I'm pretty sure that the reason people think it only works with progressive footage is that the first release only worked with progressive. However, Gunnar Thalin (the author) added the ability to work with interlaced footage, and as long as you check the "interlaced" box, it should work just fine.

The original site that hosted my guide was taken down. I wrote the guide a dozen years ago, back in 2004, and many web sites from that era have closed. However, dozens of other sites copied it. Here is one of those copies, in case you want some help using Deshaker:

http://vegas.digitalmedianet.com/article/A-Guide-to-Using-DeShaker-28849

The formatting is a little screwed up because they didn't copy the original web site very accurately, but I think you'll still be able to use it.

Deshaker has been updated several times since then, but the guide may still be useful. One other thing, while I still use Deshaker, in many cases Mercalli does a much, much better job. Of course you have to pay for Mercalli, but I don't mind paying for software if it works well and is supported by its author.

The workflow had been:

Capture to DV
Rough edit > intermediate files saved as Magic YUV (video only)
Deinterlace & resize to 960x720
Denoise
Deshake & sharpen (if required)
Final edit, render as H264/MP4 (960x720,50p) with AAC audio

If you are going to resize to 960x720, then you absolutely must deinterlace. There is no way to avoid that step when re-sizing interlaced video. However, why are you resizing? Increasing the resolution will not in any way improve the visual quality of the result. If you are going to edit video from different sources, then I can see a reason to up-res to a higher resolution, but in that case I'd go to a resolution that matched my other source material. AFIK, 960x720 is not a common video acquisition format.

... should I use your script before or after deinterlacing, I understand your explanation........... does that mean that the script "latest johnmeyer denoiser_progressive" is in fact wrongly named and should only be used on interlaced material ? or did you modify it specifically for progressive ?

When I initially responded, I said that these were my scripts. While I'm still sure that they started out as my scripts, someone may have modified them. In particular, there is no way the the script labeled as "progressive" was really intended for progressive footage. Both of the two scripts you posted use Delta=2 and Delta=4 in the MAnalyze calls, and the function is labeled MDegrain2i2. The "i2" at the end was my way of reminding myself that this function separates the video into its individual fields, denoises the even fields separately from the odd fields, and then weaves them back together.

In short, both scripts are designed for interlaced video.

They will work on progressive video as well, and should not cause any problems if used that way, but you may get slightly different results than if you used a script that simply used the entire video frame without separating into fields.

Bottom line summary:

1. Both scripts were written for interlaced video; neither was designed for progressive video, and both should be used before deinterlacing.

2. If you re-size, you must deinterlace, but you should ask yourself why you are re-sizing. Increasing the resolution does not improve quality, although sometimes upscaling can smooth "jaggies," depending on how the scaling is done. Sometimes it makes things worse. The only time I have ever up-res'd prior to uploading to YouTube was back in the day when their encoding was really bad, and by making YouTube think my video was actually an HD video, they did a better job encoding. I haven't tested this for a long time, so I don't know if that "trick" would still yield any better results.

I hope that helps. Since I can never know all the things that you are doing, I would encourage you to do your own tests on small samples of your video to see what provides the best-looking result when uploaded. As you saw above, I did that yesterday to see if uploading PAL interlaced video would result in a mess. From brief research, it appears that this was indeed a problem with YouTube back around 2011. This is not surprising since YouTube started out here in the states where most people have never heard of PAL video. However, since YouTube is obviously international, they had to quickly figure out how to deal with PAL interlaced video, and based on my one test, it appears that they handle it just fine. Of course I didn't do the additional test to see if I used some sort of advanced deinterlacer like QTGMC and then uploaded the result of that operation, whether I would have gotten a better result.

One thing I do know is that the various industries that deal with video have improved dramatically their products' abilities to deal with interlaced video. A large amount of video we watch every day on our TV sets is interlaced, and yet many sets are still natively "progressive only" devices, and they have to do their own deinterlacing prior to displaying each frame. I have not once been conscious of artifacts in all the years I've been watching on my set, and this is true when looking at other monitors. By contrast, almost every week someone sends me a video that some person has botched by trying to "improve" it. More often as not, one of the things they've botched relates to the deinterlacing operation. It is then my job (I do media restoration) to try to fix it.

This is why I try to discourage people from deinterlacing, unless they are re-sizing, and even then, as I did in this email, I try to question the reason for that re-sizing.

Two famous quotes that I live by in my media restoration business:

"Don't fix it if it ain't broke."

"Less is more."

[Tafflad]

Hi John .. reason I resized to 960x720 ... was discussed in-depth in an earlier thread ........my source is DV (non-square pixel) could have converted to 768 x 576 to square pixels, after a lot of confusion (mine) and lots of discussion here:
https://forum.videohelp.com/threads/376765-Progressive/page2
on post #58 in the thread I followed advice and aimed for output files as 960x720p, (at 50fps)

Seems I need not have deinterlaced to use DeShaker - thanks for that correction.

I would have still needed to deinterlaced as I needed to resize (to get square pixels) and as I wanted to have progressive files for Vimeo .... advice here was it is better to Deinterlace under your control rather than leave it to Vimeo/Youtube.

I will follow your advice and use your script before deinterlace in future.


On your comment "In addition, I also wrote a set of Vegas and VirtualDub scripts to completely automate the process of using Deshaker with Sony Vegas." be interested if you have any documentation (links ?) on anything you would care to share on Vegas.
I use Movie Studio Platinum ... but some of it may still be useful.

[johnmeyer]

I posted extensively in the Vegas forums about my scripts for automating the use of Deshaker. I'd dig up the link, but it won't do you any good because Movie Studio will not run scripts: only Sony Vegas runs scripts. So, I'm afraid you won't be able to use what I wrote.

I skimmed through that thread you provided in your last post and realized that I downloaded your original clip and briefly responded to you back then. However, I didn't track the thread after that, and missed all the discussion about pixel aspect ratios. My head was spinning by the time I finished looking at all the posts.

I didn't remember ever having any issue with PAR when uploading to YouTube, so I went ahead and took the liberty of uploading your knee-boarding clip to see what YouTube would do with it. Here's a private link to the result. It looks OK to me.

https://www.youtube.com/watch?v=1uOqxd-oAOg&feature=youtu.be

As for stabilizing it, forget about it. I have hours of old water skiing video, taken with VHS equipment, and you have two problems. The first is that the camera motion from the bouncing boat is pretty extreme, especially when zoomed in, and there simply isn't enough video at the margins to fill in the huge borders created by any attempt to stabilize the video. The alternative is to zoom in to correct the borders, but this makes the video look really fuzzy.

The bigger problem is that the slow shutter speed used on these early camcorders means that any time the camera moves violently -- which is about every third frame -- that motion causes the frame to be blurred because of the 1/50 second shutter speed. If you stabilize the video, each time the frame is blurred, it looks like the autofocus lost its way. This is almost as bad as the original camera motion, and it seems very unnatural. With modern cameras, if you know you are going to end up with a shaky video, you can increase the shutter speed to 1/250 or above, and this usually completely eliminates this problem.

When I played the video in the old Windows Media Player I have on my PC, as well as in VLC, neither of them seem to correctly observe the PAL 1.092 aspect ratio, and as a result displayed the video so that it was squished horizontally. I don't know if that happens on your PC and, if it does, whether that contributed to some of the confusion over non-square pixels and how it is handled by YouTube, Vegas, WMP, etc.

[Tafflad]

>My head was spinning by the time I finished looking at all the posts.

I'll admit much of it confused me ... especially when what I was reading in Mediainfo and the like - I was told was actually wrong as wrong values written .........

However I could follow the logic on square pixels and going to a standard HD size for the streaming files. .... hence 960x720p

> neither of them seem to correctly observe the PAL 1.092 aspect ratio
Didn't have that problem ... all player I have play ratio OK .... VLC does not seem to know when its interlaced though and have to manually set.

I tend to be selective with DeShaker - if output looks better I use it ... not always the case.
I originally had advice to set Motion smoothness to 1000 or greater (H,V,R & zoom) but then advised to look at 2-4 x the frame rate and leave zoom to zero - and it is much better.
I tried "use previous and future frames to fill in borders" (and "soft borders") but had no success, just a lot of mushy edges ... maybe due to the conditions you described above.

[johnmeyer]

I know what you mean about stabilization sometimes making everything look worse. Some of that is due to the nature of how the process works, and some of it has to do with deficiencies in Deshaker. It is a wonderful piece of work, but it does have some limitations. The border fill is a great feature, but as I'm sure you have experienced it often fills in with some pretty weird video.

If you have the time and inclination, you could download the ProDad Mercalli demo and try it out on a few clips. The standalone version has more features than the plugin. It is very expensive, so you may not have the budget to buy it, but at least you'll know if something can do a better job on your clips. If you like it, maybe someone will give it to you for Christmas!

[lordsmurf]

Too many VH threads have arguing these days. Reminds me of Doom9, Afterdawn, and some other now-gone (good riddance!) sites.

My input on a few items:

- I've never liked Deshaker. In fact, I think it's outright crappy. Several have tried to make guides for it (trevlac, John, others), but at the end of the day, it's just an awesome guide for a crappy program.

- Some Avisynth stab methods work, but only for minor issues.

- Mercalli isn't expensive to me. It's no worse than something from Adobe, MainConcept, and others. I have it, and it's easily the best stabilizer for non-minor needs. (Again, thanks John, for tipping me off to it.)

- I'd deinterlace here too.

- John's expertise is film. Some of his filters are going to yield bad results on VHS sources. And no, it really does not matter whether you deinterlace first or not. Film vs. tape is more than progressive vs. interlace. Anytime I need to work with film and Avisynth (rare), I find myself perusing his past scripts for tips. I've seen a lot of arguments and confusion over the years because of this. You can't just apply his methods to all video.

[johnmeyer]

Too many VH threads have arguing these days. Reminds me of Doom9, Afterdawn, and some other now-gone (good riddance!) sites.

Mea culpa. I got tweaked and then responded. I should know better. My fault.

I've never liked Deshaker. In fact, I think it's outright crappy ... Mercalli isn't expensive to me. It's no worse than something from Adobe, MainConcept, and others. I have it, and it's easily the best stabilizer for non-minor needs.

I'm not sure I'd classify Deshaker as being that bad, but Mercalli most definitely blows its socks off.

John's expertise is film. Some of his filters are going to yield bad results on VHS sources.

Well, I do actually do video as well as film, but if someone uses anything that I developed for film, expecting to get good results with video, they will most definitely be sorely disappointed.

The VHS denoising script that I've posted many times, and which was the subject of some of the discussion earlier in this thread works pretty well. That is not just my own opinion, but has been the feedback I have received in numerous other posts where I have offered it as a starting point for consumer home video restoration. It is absolutely nothing unique or special and is pretty much cribbed from the MVTools2/MDegrain2 documentation, although it does include some code to help with 2nd generation color offset problems.

And no, it really does not matter whether you deinterlace first or not. Film vs. tape is more than progressive vs. interlace.

Yes. One issue is the nature of the "noise" (film grain is not the same thing as video noise). Another is the chroma artifacts (film doesn't usually have any, unless they are introduced by color conversions, or some error in the process). One of the biggest issues is temporal differences. 24 events per second can give you huge spatial jumps between frames, especially with fast moving motion, compared to 50 or 60 events per second for video. The problems introduced by these big jumps get much, much worse when transferring amateur silent film which often was shot at a speed as slow as 12 fps (old 1920s film that I've transferred), but no more than 18 fps. One example of a restoration technology that doesn't work as well at films speeds is motion estimation: it barfs (technical term) at these big jumps, and often fails pretty badly.

[Tafflad]

If DeShaker is that bad ......... are there any other better free options. ?

I did take a look at Mercali web site .... all their sample videos are for slow moving camera and equally slow moving or static objects ..... may not work at all for my videos ........... I might try downloading the demo version and try it.

[johnmeyer]

If DeShaker is that bad ......... are there any other better free options. ?

DepanStabilize.

However, don't take anyone else's word about something being good or bad. Have you gone back to Deshaker, with the interlaced option checked, and tried it out again? What was your experience? It works quite well on many clips, but like any tool, it does fail on some things.

[Tafflad]

Don't have any interlaced projects to try yet .... maybe I'll try out some tests soon.

[lordsmurf]

DepanStabilize = too many false positives

Only very minor settings work well, such as the stab() or my stabmod() implementation.

I tried for years to get it to behave. I wish I'd have known about Mercalli years ago.

(Then again, only in v3 and v4 has it's been useful. And BTW, if you get it, get the SAL, not the plugin. The export format choices really suck, but MP4 output is passable.)