Deinterlacing Problem - QTGMC & NTSC Footage

[hickeyguy]

Good evening, everyone!
I'm very new here, and also pretty new to the world of deinterlacing. For a couple of years now, I have been trying to correctly deinterlace some DVD footage that has been the most difficult thing to figure out ever, out of all of my DVDs. When I first started, I used Wondershare software to simply cut each clip I wanted to extract, and it would deinterlace automatically, BUT would make the video laggy (repeated frames). Then a while later, I came back to the project, and found out how to stream VLC deinterlaced footage (using VLC's range of different deinterlace options) to file. Which gave much better results! BUT far from perfect. I've always wondered if there was maybe a combo of Telecine & Interlace happening..... but after recent tests, I think it is just Interlace that I need to find out how to deinterlace properly. So recently, after such a long time not playing with it (because it is so frustrating), I found a deinterlacer called QTGMC. Looks very complicated, but I found that StaxRip has this plugin built into its GUI. So yesterday I got back into trying to deinterlace my DVD footage. Just a moment though - for your information - the DVD I used originally a couple of years ago was a PAL DVD. I bought a collection of the same series but in NTSC last year, and did a side-by-side comparison to see which had a better image quality (pre-deinterlace and post-deinterlace), and the NTSC did indeed look better overall, even with a few less pixels (being NTSC). The NTSC version also gave me slightly better results when using that VLC deinterlace trick - but still problematic as it either gave me lag on some options, or remaining interlace/telecine lines on other options. So I gave up for a while. Anyway, back to yesterdays QTGMC! QTGMC via StaxRip has given me the BEST result yet, and i'm very happy about that (as I remember having really poor quality using handbrake a year ago or so, as a test, as I wanted to see how other deinterlacers worked).

Problem number one I have with everything that I would really appreciate help with is figuring out the pixel aspect ratio of the original clip, so that my rendered clip, doesn't change to a different size (my renderings squash slightly, and I can't figure out what the correct setting is - pixel aspect ratio?). THe resolution from Original clip to rendered/encoded clip are the same 720x480, but squash downward on output. Problem number two I have and need help with majorly, is getting QTGMC to deinterlace correctly. It gives me an amazing result - zero lag and no interlacing lines in areas of motion, which is thte best i've seen yet! Fantastic. However, on objects, and solid lines, like a telegraph pole electricity wire, or cats whisker for example, will not appear as a straight line, but as a zigzag. The clip I will attach here, will show exactly what I mean. If you use QTGMC on the attached clip (extracted from NTSC VOB's using DGIndex), it will remove the interlacing (and also double the frame rate to 59fps/60fps), but in the result, you can see the bottom of the van is jagged, and not a clean straight line (moire? or incorrect deinterlace?). Something isn't working right.

Please try to help me with this, as it has been driving me nuts these past couple of years. Still can't achieve a 100% deinterlace for this footage. Will appreciate any QTGMC sample codes (which can be added in via the StaxRip filters options -for Field>QTGMC). I reckon i'm doing something wrong, and would absolutely love it if I could learn how to do this properly, after all these years. Any help as to fixing this problem, I will greatly appreciate! Thank you!
Kind Regards From Guy!

[manono]

Too much information. Yes, you want to provide everything necessary to those willing to help, but that was way too wordy. Also, it's a very poor sample. Please provide one with steady movement, not just some guy squatting and moving his arm a little bit. Yes, you wanted to show the aliasing. Actually,I think the content is progressive 29.97fps, meaning you don't need to bob it with QTGMC.

The reason for the aliasing, like along the side of the van, is because there's not enough resolution to resolve fine lines. It has nothing to do with interlacing. It's just a crap DVD. It's a widescreen source that should have been encoded as 16:9, not 4:3. All that black above and below should have been filled with picture.

You might want to try out an anti-aliasing filter.

[hickeyguy]

Thank you so much for such a quick reply, Manono! Very new here, please allow me time to adjust to how this place works. I want to learn how to do this properly, once and for all. I uploaded that clip because it is the one area of footage that I could quickly share, that shows an area of problem. If that source clip (attached above) is deinterlaced using QTGMC (the best I could find thus far), it leaves the following ugliness which I would love to figure out why it exists exactly and how to fix it =

[Attachment 48168 - Click to enlarge]

Is the QTGMC Progressive Input Mode something that would mean that I use QTGMC to deinterlace the source footage, and THEN use QTGMC again but on the 'deinterlaced' footage? (so, a second pass of QTGMC? In this order = Source>QTGMC = DeinterlacedFootage..... then DeinterlacedFootage>QTGMC with ProgInputMode code added = Deinterlaced fixed footage?).

Thank you. I would love to see if anyone might experiment themselves (with their valued expertise and skill), to see if there are any methods that will 100% work on this clip. Maybe something other than QTGMC might produce a better result?

Many thanks & looking forward to any conversation this might bring!

Edit = Just saw that you edited your response. Wasn't expecting that - Forgive me for being under-knowledged with all this video technical stuff. But if this is antialiasing(?), would you happen to know what process I might need to follow in order to call on AviSynth and an antialiasing filter via StaxRip? Or maybe a quick overview of how you might do this process yourself, personally (step by step, briefly)... I'd love to learn how to get a 100% quality output from this source media. Willing to stick around and learn if someone here has some patience with me. Thank you!

[JVRaines]

The file you provided is not truly interlaced. The jaggies are baked into the image. It is at 10:11 pixel aspect, typical for 720 × 480 sampling of a 4:3 frame.

[manono]

Is the QTGMC Progressive Input Mode something that would mean that I use QTGMC to deinterlace the source footage, and THEN use QTGMC again but on the 'deinterlaced' footage? (so, a second pass of QTGMC?

Yes, I edited. I tried out QTGMC's Progressive Input Modes and they didn't seem to help much. So I edited to point you to other anti-aliasing filters. I didn't try any of them. Anti-aliasing might also degrade the rest of the video.

As JVRaines also said, it doesn't seem to be interlaced at all. The 'jaggies' are part of the picture for reasons I already explained.

[hickeyguy]

G'day JVRaines! Thank you for your quick reply as well. I find it very interesting that the jaggies might be there, baked or pre-interlacing from the people who made the DVD's (meaning that they will always be there on both PAL & NTSC sourced DVD's). Does this then mean that the following lines (in the following screenshot/vlcsnap) are possibly telecine lines? (screenshot from a different area of video containing lots of motion, compared/opposed to the van clip from above, which doesn't really contain any motion for the interlacing to appear... just those annoying jaggies or antialiasing?) =

[Attachment 48169 - Click to enlarge]

And thank you for your observance on the pixel aspect ratio. I'm confused as to why StaxRip squashes my output, when PAR, DAR & Resolution are all the same (maybe the SAR number is ruining it for me?)... here is an example of what I mean (yellow being where the squash occurs) =

[Attachment 48170 - Click to enlarge]

Would love an example process (step by step) from someone that is able to properly deinterlace or detelecine or fix this problem i'm having. Happy to provide more DGIndex clips from the source NTSC VOB's. Please do let me know.
Thank you!

[hickeyguy]

Thank you, to both of you!

Here is a second clip extracted with DGIndex, which shows both the interlacing (or telecine?) and the jagged/anti-aliasing attached as filename [1a-Source.m2v].
I have also attached [1b-QTGMC.mkv] which StaxRip has produced. From what I am assuming, it has deinterlaced with QTGMC wonderfully with no resulting lag (repeat frames) or any remaining combing of which I have received from other deinterlacing methods in the past (using Wondershare software, Handbrake, and VLC's built-in stream-to-file options). The only problem I see (a part from the doubled framerate and slight squash), is with those ugly jagged lines on the truck, which if you are right 'is baked into the original video to begin with'. Here is a screenshot of my StaxRip job settings =

[Attachment 48176 - Click to enlarge]

I would like to know why my output is squashed slightly - as DAR, PAR & Resolution are the same. SAR isn't editable, from what I can see... is that the problem? (regarding the squash?). The only other thing i'd like to figure out is how to fix those jagged lines... there has to be a fix for that, even if it is in the prior progressive master (pre-interlaced-DVD).

Thank you, all, for your time! Looking forward to seeing how we go with this. If there's anything I might be able to do to contribute in return, for your great help, i'd love to know. Much appreciated!

[manono]

I find it very interesting that the jaggies might be there, baked or pre-interlacing from the people who made the DVD's (meaning that they will always be there on both PAL & NTSC sourced DVD's).

No. Didn't you understand what I wrote before:

The reason for the aliasing, like along the side of the van, is because there's not enough resolution to resolve fine lines. It has nothing to do with interlacing. It's just a crap DVD. It's a widescreen source that should have been encoded as 16:9, not 4:3. All that black above and below should have been filled with picture.

If there's a decent 16:9 DVD out there of it - NTSC or PAL - then no more aliasing (probably)

Would love an example process (step by step) from someone that is able to properly deinterlace or detelecine or fix this problem i'm having.

That pic does show interlacing, plus some kind of chroma aberration. But without a fresh sample showing the problem, no one can help.

Edit: Now I see you uploaded a second clip. No MKVs. M2Vs are preferred.

[hickeyguy]

Thank you, Manono. Unfortunately the PAL & NTSC DVD's i have are the only one's out there that's available. No 16:9 releases, although I haven't been able to see what the iTunes versions of this series are like on the WEB (probably a lower bitrate though - probably progressive as well).

That MKV was my QTGMC render through Staxrip (labeled 1b-QTGMC.mkv). The M2V is the source (1a-Source.m2v).

[manono]

The second one is telecined. It just needs a simple IVTC:

TFM().TDecimate()

If your video is a mix of progressive 29.97fps and telecined film, then it might be best just to bob the whole thing.

And I was rethinking the first one. The sample wasn't good, but there's a chance it had been poorly deinterlaced, and that may have created the aliasing. Along with creating a 4:3 DVD when it should have been 16:9.

[hickeyguy]

Thank you, Manono. I had a feeling telecining was in there somewhere. I've had much confusion between the two (telecine and interlace), regarding these WBz DVDz over the past couple of years. I think the PAL version I have has a framerate issue (regarding interlace or telecine, or both), maybe in relation to it being converted from an NTSC source (possible?)... This NTSC version I have been experimenting with lately has given me better results than the PAL version, but as it appears, there might be a problem with the original master material used to create these DVD's in the first place. It looks like i've got a lot more experimenting and exploring ahead of me! I have to figure out a way to remedy those jaggies where lines should be straight and not zigzag'd. Thanks to you, Manono, I will now explore the Anti-Aliasing route (something I haven't looked into before), to see if I can get a result that corrects this problem (even if a little degrade occurs overall).

To anyone new to this thread - please do have a go and play with the '1a-Source.m2v' file I attached in my 2nd last post - it is the file I will be using for all of my next tests. Fingers crossed I find something that works, as i'd hate to have to keep on playing with this forever.

[hickeyguy]

Thank you, Manono! I went ahead and turned off QTGMC in StaxRip, and just applied a 'simple IVTC', and it yielded a much better result!!! I have to explore it more, to see if I get any combing elsewhere (like in any of the motion graphics - a problem I has having late last year using VLC's IVTC stream output), but I am liking where it's all going. Have a look at this comparison I put together =

[Attachment 48177 - Click to enlarge]
Thank you, again!

Edit =
Maybe a little more tweaking is needed somewhere. I have to look for codes that might help with the IVTC. It's almost perfect, but as you can see in the below animation, one frame every so often, shows jagginess (or antialiasing?). I have also noticed that IVTC automatically changes the 29.97fps source down to 23.976fps in the output render... maybe IVTC is having trouble with the motion? (see jagged frame in animation below) =

[Attachment 48178 - Click to enlarge]

Edit 2 =
I do have to give you thanks, yet again, Manono... I was trying to figure out how to make the codes work in StaxRip, calling on the specific code you mentioned. SO I tried to play with the only option I had access to, changing the IVTC Filter code to =
TFM().TDecimate()
Instead of the IVTC default code =
Telecide(guide = 1)
Decimate()
So I figured that if you replace the default with the new code, it should either give an error, or work? And it did! I'm very happy with this last result! Here is a before after comparison, showing the default IVTC code result, and then your TFM code instead (much better result!) =

[Attachment 48179 - Click to enlarge]
Thank you. Now I just need to put this new method to good use in an experiment, and encode an entire 20min episode, to see how it all turns out. The only other thing I do also have to play with, is the PAR, to try and get it to NOT squash, like it has been doing (for a reason I am yet to learn about).

Edit 3 =
Thank you again, Manono! I was searching for answers as to why the framerate drops down from the original source down to output when using the IVTC filter, and your comment HERE along with material I found reading about what Inverse Telecine is, that was all I needed. Thanks! Fingers crossed I create a great render tomorrow! (this time I will output to MP4 instead of MKV, which might be more compatible in the long run).

[jagabo]

After IVTC add

Blur(0.0, 0.5)
Santiag()

[JVRaines]

The only other thing I do also have to play with, is the PAR, to try and get it to NOT squash, like it has been doing (for a reason I am yet to learn about).

Try 8:9.

[hickeyguy]

G'day jagabo & JVRaines!
Thank you for your added input here. As for the aspect ratio problem I was having, I managed a way around it by doing something that may be frowned upon here, but for my test it worked just as I wanted it to (thankfully). I changed the 'height' from 480 to 540 (StaxRip had me change a setting called "Output Mod" from number 8, to number 4, so that the height could be divisible by that number - something 8 & 540 weren't compatible, but 4 was). I didn't mess with the automatic PAR, DAR or SAR (just the resolution). I will definitely do a test later today and try your suggestion, JVRaines - thank you!

For you code, jagabo - I would like to try a test to see what your code does, but I think I need you help as to how I need to implement it into the workflow i'm using (with StaxRip). Do you know how I need to enter this code, as placing it in the IVTC field, after manono's IVTC code, gives an error (seen in screenshot below). If I were to add another entry to the script/codes filters list, what might I add (to enter your code - like some kind of 'custom code filter' where that code can be placed in the list?). Thanks!

[Attachment 48185 - Click to enlarge]

[jagabo]

You're have the code in the right place but you need to download the Santiag() filter.

http://forum.doom9.org/showthread.php?p=1393006#post1393006
http://avisynth.nl/index.php/Santiag

Copy the text and paste it into a file called Santiag.avsi. Put that file in AviSynth's plugins folder and the filter will be available any time your use AviSynth.

[hickeyguy]

Thank you very much, Jagabo. I did as you have instructed, and successfully got a render after I placed Santaig.avsi (with the script in that file) into the AviSynth+ plugins folder (along with nnedi3.dll - as it was required). The following image is a comparison between the IVTC code Manono suggested and then that code again but with the added Jagabo code as well =

[Attachment 48187 - Click to enlarge]

You can see that the blur in the code does make it blurred slightly (details in the grass, tail light and tire for example, have degraded). Is this the Anti-Aliasing degradation from the filter that Manono mentioned before? If so, I think I might stick with TFM().TDecimate() and not TFM().TDecimate() Blur(0.0, 0.5) Santiag(). Thank you though! I've learned something, and I really appreciate that.

Right now, after a day of experimenting and contemplation, I am really happy with how the 'simple IVTC filter' processed the video, compared to the QTGMC filter which was unnecessary. The only problem I have with the IVTC right now (which I want to see about fixing somehow) is the framerate. Both the source and QTGMC renders run very smoothly, as the source is 29.97fps and the QTGMC is 59fps (motion is so smooth and fast). The IVTC takes it down to 23.976fps, where I can actually see very slight stutter in motion. Maybe stutter is the wrong word... It just isn't as "smooth" (to my eye) as the videos with the higher fps. I would love to figure out how to remedy this!

Does anyone have any idea's as to how I might get this IVTC filter to give me this amazing detelecining result, but keep a higher frame rate? (so I have the smoothest motion possible). It is only very slight, but enough to bother me. That would be amazing if we could fix that!

Thank you!

Edit =
I wanted to illustrate what I meant by this difference in framerate giving me a slight degrade in the motion. I have attached an MP4 clip here which I created in After Effects, placing my QTGMC 59.94fps file in a comp, with my 23.976fps IVTC file ontop > then splitting one down the middle with a mask (pictured below) to show "both halves" together. The 59.94 fps QTGMC version is ON THE LEFT (and plays smooth)... and the 23.976fps IVTC is ON THE RIGHT (and plays not as smooth, which bothers me). You can see that the RIGHT side isn't playing as smoothly as the LEFT. That in my mind, has to be because of the loss of frames through the IVTC filter dropping the render down to 23.976fps (does that sound right?). I'd love to figure out how to get the same amount of information/frames that would exist in the source telecined 29.97fps file > and get it all into a smooth detelecined/IVTC result (so that all the motion is smooth like the 30fps source and 60fps QTGMC render). I hope that makes sense, and I hope you can see what I mean. At points in the comparison SideBySide video, they sync up, but at points you can see that frames are missing in the 24fps side (right) and is jumping slightly, causing a loss in smoothness of motion.

[Attachment 48189 - Click to enlarge]

Edit Yet Again =
I have tried something else, digging in a little deeper to inspect, and try to see if what I was suspecting, even made any sense (my brain has so much info running through it - please have patience with me). If the master footage is assumed to be 23.976fps (just assuming), then it would have been converted into 29.97fps for NTSC DVD purposes (telecined? Not interlaced). Then using the IVTC filter would bring it back to original FPS for a reason (Inversed). So, with that in mind, I checked my 59.94fps render from one of my test clips, frame-by-frame, and noticed that the 59.94fps QTGMC version had either 1 or 2 repeat frames for every SINGLE frame (with some morphing in those doubled frames), meaning for every source frame, it doubled or even tripled (maybe QTGMC morphed between frames creating 2 extra frames?). Anyway, in my 23.976fps IVTC version, the exact same frames are there, minus the doubled or tripled frames seen in the 59fps version. So maybe it is just a trick of the eye! (with the lag in motion I was seeing in the attached video). The motion is as smooth as it is going to get, if there are only so many frames to begin with. The progressive FPS has to be 23.976 at the end of the day. Does everyone think this is correct, in this case? I need to know for sure. Thank you very much!

[jagabo]

The point of adding Blur and Santiag was to reduce the aliasing on near horizontal edges that are visible in the "after" shot in post 12. It reduces the "buzzing" of those edges when viewed at normal playback speed. Of course, the result is less sharp. If it blurs too much for your taste lower the vertical blur value or remove the Blur entirely. Or leave out both and live with the buzzing edges.

The source you provided was shot on film at 24 frames per second. It was then slowed to 23.976 frames per second and telecined to 59.94 fields per second analog interlaced video. When digitized, pairs of fields were woven together to create 29.97 frame per second digital video. IVTC with TFM().TDecimate() reverses the process and restores the original 23.976 frame per second progressive film frames. That is the correct way to handle this video (unless there are native 59.95 field per second shots mixed in with the telecined film). When this 23.976 fps video is displayed on a 59.94 Hz display the frames are repeated in a 3:2 pattern. That is, one film frame is repeated for a total of 3 frames, the next for 2, the next for 3, the next for 2, etc. So 23.976 frames becomes 59.94 frames (or fields in the case of old analog interlaced TVs).

QTGMC peeled the digital interlaced 29.97 frame per second frames apart into 59.94 fields per second, then attempted to turn each field into a progressive frame using information from fields around it. The result is 59.94 frame per second video with repeats of the original film frames in a 3:2 pattern. Its deinterlacing isn't perfect so some the repeats may not be exact duplicates. There may be some small defects where detail is lost, horizontal edges bounce a little, etc.

[hickeyguy]

My word! Jagabo, your reply is as perfect as I could possibly hope for! I'm about to go to bed, and want to absorb everything properly tomorrow - but have to write now to give a "thank you!". This is the kind of information I really appreciate - detailed enough to explain what is going on so I can understand in my mind (almost visually, in stages). It's so fascinating! I will try your code again a some point, adjusting some of the blur (or removing the blur altogether) just to see what happens. Maybe some post-blur sharpening might be on the horizon somewhere, as I did read about something in a recent post here about SD to HD upscaling (regarding some sharpening as opposed to Super Resolution, or something like that).... A lot of me to think about going forward!

Thank you!

Edit =
Today I have spent a lot of time inspecting some of my renders FRAME BY FRAME to really understand why I was seeing that "loss" in motion smoothness last night. I thought it was just a trick of the eye (something psychological, knowing the fps is lower), being 23.975fps. But after some reading and an experiment, I have figured out something really strange that makes no sense to me (as I believe the information everyone is giving me - I do not concur with anyone here, but in this case something is different). I went ahead and rendered a file removing .TDecimate() from TFM().TDecimate() (to stop the removal of the assumed duplicate frames), and then reviewed SIDE-BY-SIDE and FRAME-BY-FRAME to see if there were any duplicate useless frames (like there was in the QTGMC render at 59fps, containing up to 3 extra interpolated? frames). I was surprised to find that not only did this new code detelecine the source just how I wanted (thank you, Manono!), but it contained no duplicate frames, and did not feature any "missing frames" like the 23.976fps file did indeed exhibit! I was shocked, but happy... and confused. But satisfied with this detelecined file at the frame rate of 29.97fps.

That doesn't make much sense, does it (detelecined and keeping the same framerate as the source file).... But maybe this was never a 23.976fps film master to begin with? I just can't explain how there aren't any missing frames or duplicate frames, in either method/process. Is this news to anyone here, or has that happened on some NTSC DVD's before? I even had a look into SVPflow to see if there was a way to create fake interpolated frames, just for fun! But it produced some ugly stuff, in area's of extreme motion. Maybe that code I used needed a bit of an adjustment.... but i'm happy that I found this solution in the end. Tomorrow I want to go over everything I have explored in the past few days, just to double-check where i'm going, as I have a ton of data to go through (to detelecine), and no room for making mistakes now before I go ahead with that.

[hickeyguy]

To add to my last post edit - I have created an animation using the extracted frames from those two clips I rendered (one with just TFM, the other with TFM & Decimate in the code - for the IVTC filter).

On the LEFT is the 23fps IVTC with TFM&Decimate.... On the RIGHT is the 29.97fps IVTC with TFM but NO Decimate =

[Attachment 48212 - Click to enlarge]

The LEFT (with Decimate) is missing frames (and the RIGHT has the original amount of frames), and it isn't a trick of the eye as I had previously thought! I had a feeling something funny was going on with the motion... I'd love to understand why this is the case?

[jagabo]

That's not taken from the clip you provided in post #7. As was pointed out several times, if other parts of your source (or other sources) aren't telecined film they will require different treatment. Provide a sample of that new clip if you want analysis of it.

[hickeyguy]

If your video is a mix of progressive 29.97fps and telecined film, then it might be best just to bob the whole thing.

I believe this may be the case, as the original clip I attached at the beginning of this thread (with the van) still shows the 'major jaggies', with or without the TFM treatment. As of last night, I thought we finally found the solution (after TFM was able to fix all the other clips I was focusing on), but I think the episode is a mix of progressive and other things. TFM works great in most of the video, but in this 'van part' (from post #1) the jagged lines are going crazy. So there must be a mix of some kind going on, and no solution for it (I tried every Field filter in StaxRip with no good results - maybe the lines are just too small to fix, like manono mentioned before "not enough resolution to resolve fine lines").

Thank you, Jagabo - I wish I could upload the entire 923MB VOB for you! Not sure what to do now. I might need to sit on it for a while, and come back in the future sometime when I can focus better.

[jagabo]

'van part' (from post #1) the jagged lines are going crazy.

That clip was original interlaced video (59.94 fields per second). It was very poorly deinterlaced to 29.07 fps progressive, then encoded 29.97 fps interlaced. It's progressive frames with 29.97 unique pictures per second. TFM() will not improve the picture. TDecimate() will make it jerky. QTGMC(InputType=3) will help a little.

[hickeyguy]

That clip was original interlaced video (59.94 fields per second). It was very poorly deinterlaced to 29.07 fps progressive, then encoded 29.97 fps interlaced. It's progressive frames with 29.97 unique pictures per second. TFM() will not improve the picture. TDecimate() will make it jerky. QTGMC(InputType=3) will help a little.

G'day Jagabo!
Thank you. I actually had a look through torrent search engines, to see if anyone out there has ever successfully deinterlaced/detelecined (or whichever) these DVD's, and I found a Russian page which had MKV encodes for all of these episodes including a MediaInfo breakdown for the MKV's (not sure if PAL or NTSC though - I believe NTSC - I have both in my collection, but both have proven to be very problematic with these jagged lines). I downloaded them, and noticed that they have had the exact same problem i've been having (the jaggies are everywhere, in so many scenes, throughout all 4 seasons!). So it isn't just me that has had trouble, I see - which makes me feel a little better, as i've been feeling like i'm missing something. I had a go with your QTGMC(InputType=3) code suggestion (thank you! - also tried type 1 and 2 along with ProgSadMask to see what that would do), and did not see ANY change regarding the jaggies in the m2v clip shared in post #1).

I would absolutely love and dearly appreciate if you or someone could have an experiment with that clip (in post #1) and try to see if you can do anything to affect those jaggies. I've tried so many deinterlacing filters, and different softwares (tried Dscaler last night with some difficulty), and nothing is working on my end. The only glimmer of hope i've seen, is when using VLC, starting with the video PAUSED - when I TURN ON the deinterlace function (with any random deinterlacer selected as default - ANY of them produce this weird result) the image loses quality (blurs slightly), but there are no jaggies to be seen!!! BUT, the moment I pres play or 'E' to go to the next frame, the jaggies appear again, and that frame never appears again unless I turn off the deinterlace function, and turn it back on again (with the video paused). Here is an example framegrab of the moment I turn on the deinterlace function.... it only seems to last for 1 single frame, and it seems to appear that the deinterlace no longer affects the footage after resuming to further frames in the video.

[Attachment 48237 - Click to enlarge]

And for your reference, this is the next frame - and as you can see the deinterlace that VLC applied to the previous/above frame, no longer is in effect =

[Attachment 48238 - Click to enlarge]

Surely there is something in AviSynth/StaxRip that can fix this. I am highly intrigued by this! If it is baked into progressive frames (or whatever the case), i'd love to learn how to manipulate it in a way to stop them from appearing. I'm sorry if a solution has already been shared with me (if anti-aliasing, as mentioned previously, is the answer?), but i'm still new to this, after years of starting/stopping this project of mine. I feel like an annoyance to everyone here, but I was expecting this to be a challenge people might enjoy learning from along with me! If more m2v's are desired, please do let me know - however, I think if the clip in post #1 can be fixed, that there is hope for the way forward through the rest of the videos.
Kind regards!

[poisondeathray]

I'm sorry if a solution has already been shared with me (if anti-aliasing, as mentioned previously, is the answer?)

Yes, you can use an antialiaser (AA) as mentioned earlier. There are dozens to choose from; some are better at different types of aliasing (or some settings work better for specific types of aliasing patterns)

In general, AA filters are very damaging to fine details, so in general you'd only want to use them sparingly (not that you have much fine detail here to begin with, so you can probably go overboard)

Some work intra frame only, some are temporal. You usually need some temporal smoothing for video, or you will get flickering in motion across frames

You might use different settings, or different types of AA filters for different scenes. You might even apply them selectively through masks (apply to different sections of an image, selectively) as you don't want to blur or damage all the details to much

Your two samples have different characteristics, so you 'd have to treat them differently (this one is actually progressive and badly deinterlaced)

For this scene ("WBz_S1_D1_EP1_South_Africa_1_RAWclip.demuxed.m2v ") , it might look like

MPEG2Source()
EEDI3(field=1)
QTGMC(inputtype=1, sharpness=0.2, preset="very slow")

[hickeyguy]

G'day poisondeathray!
Thank you for your detailed response. Very much appreciate it! I just gave your script a go (after placing eedi3.dll in the plugins folder), and have just received the BEST result i've seen yet, for that clip shared in post #1. I'm shocked, as i've tried so many deinterlacers - but this script you have kindly shared with me, has given a really nice result. Here is a screenshot, which i'm very excited to share =

[Attachment 48240 - Click to enlarge]

And my goodness! I've just compared it to the above 2 screengrabs I shared yesterday, and it even surpasses the VLC deinterlace that I was so confused by. Wow!

I have some further testing to do tomorrow, and some reading to do for EEDI3 (I need to learn more about it). Exciting stuff, poisondeathray! Are you a fan of the MTV series Wildboyz? Well, from one fan to another (even if you're not), thank you so much. Fingers crossed with this new start, I can figure out what to do next after tomorrows tests. I want to encode the entire episode just to see how it affects other scenes. I am also very happy with this script as it keeps the original 29.976fps... I was worried that QTGMC might interpolate it up to 59.94fps, but it looks like it didn't .

Look forward to updating everyone soon.

[jagabo]

QTGMC is causing the guys fingers to disappear at times. You might try leaving it out or reducing its settings. Of course, there will be less smoothing of the buzzing edges if you do that. Find the balance you're comfortable with.

[hickeyguy]

QTGMC is causing the guys fingers to disappear at times. You might try leaving it out or reducing its settings. Of course, there will be less smoothing of the buzzing edges if you do that. Find the balance you're comfortable with.

G'day Everyone!
Just a quick update from me today! I've had a lot of fun - yesterday I spent a good amount of time testing and comparing the quality of pixels in frames (between each encode), and i'm very happy with it so far. The below is a quick animation showing 4 frames - 1 original M2V, 2 EEDI3 & IVTC/TFM(), 3 EEDI3 with the suggested QTGMC code, and 4 an upscale with added noise and sharpen (maybe too much sharpen), just for fun (to see if I could improve the picture quality in Premiere Pro).

[Attachment 48258 - Click to enlarge]

Today, I have tried searching around for reading material. I've only scratched the surface regarding all of the documentation and forum threads about EEDI3 & QTGMC, but I found that adding NoiseProcess=2, NoiseRestore=1.0 seemed to bring back a lot of the detail that QTGMC removed by default, but didn't really affect the magic EEDI3 did on those jaggies thankfully!). I haven't created an animation or side-by-side comparison for everyone to see, but it really did the trick. Still more testing for me to do, as I still need to try this noise bypass thing on another scene.

Thank you again, Everyone! I just love where we've reached thus far. I'm not sure if there's really anywhere further to go from here. I will definitely add more here in this thread, if I find anything or need your assistance. Actually, just for fun - i'd love to know if anyone has an AviSynth process for upscaling, like Infognition's SuperResolution plugin, but better? Maybe that's not really possible right now - creating detail where none exists (maybe something for in the future - as I had a read of an article about GAN 2.0 - NVIDIA’s Hyperrealistic Face Generator, and about how these learning computers could be used for super resolution - maybe better than using algorythm's like Steffen Gerlach's 'A Smarter Scaling' or Topaz Gigapixel).

Kind Regards!

[manono]

i'd love to know if anyone has an AviSynth process for upscaling...

NNEDI3_Rpow2, perhaps with some sharpening afterwards.

A sample script might go:

nnedi3_rpow2(rfactor=2,cshift="Spline36Resize",fwi dth=704,fheight=480)###VCD to DVD

[hickeyguy]

i'd love to know if anyone has an AviSynth process for upscaling...

NNEDI3_Rpow2, perhaps with some sharpening afterwards.

A sample script might go:

nnedi3_rpow2(rfactor=2,cshift="Spline36Resize",fwi dth=704,fheight=480)###VCD to DVD

Geez louise! Thank you, manono! I had a go (and did a little quick reading as well) and made a comparison between my fun test last night (using Infognition's Super Res plugin - for AFx/PremierePro) and tonight's NNEDI3_rpow2 code you have shared with me for AviSynth.

I amazed, as after my initial comparison, I didn't think it worked very well. But after a closer look, I am amazed by the result! Have a look at this 2 frame animation/comparison! =

[Attachment 48260 - Click to enlarge]

Isn't that amazing! More detail AND the colours haven't altered (like it did through the Super Resolution plugin). Both of those frames are screenshots from within After Effects, so the comparison is true. Thank you manono - I appreciate it very much, and will be reading more about this going forward. My plan is to first create EEDI3-QTGMC renders for all of the episodes using all of the information we have shared and learned about here, over the past week - and then afterwards, try to upscale to 720p, as a kind of fun project to see if I can improve the image quality (as HD versions of this show have not yet been released on Blu Ray, or on WEB streaming services). That's pretty exciting!