Greetings all and a happy new year...

I have finally finished the editing of a video project I've been working on for...something like 15 years. The project was shot in standard definition because that was all that was a round when I started acquiring gear etc. Seeing I did all the filming of myself as a talking head, I could not used optical zoom to create variation, and have used digital zoom in editing to create this. However the more you zoom in the more the pix-elation shows.

This is something I may have to live with, but I thought that I may try up-resing the video, if possible, once it was all finished, and am at that point now.

I have come across:
Video Enhancer: resize video with Super Resolution method from SD to HD (http://www.infognition.com/videoenhancer/)
which purports to to be able to do this, but I was wondering if anyone has any experience in this process, or knows of other programs to do this, and perhaps could guide me in regards to it.

Thanks

Renton

Couldn't get that website to open - which does not give a very good impression of the company... But unless it uses some kind of very advanced AI to create detail that was never there in the first place, I'd question the value.

Have been gradually working with footage of opera rehearsals and performances that I shot on my VX-2000 in 480i60 back in 2002. For a 1-hour documentary, I did not up-res anything, but de-interlaced it with Compressor on my Mac and converted it to uncompressed 480p30. Did a lot of effects work on the video, so I thought this would be the best approach. Looks pretty good - had forgotten how nice the VX-2000 was for its day.

Did another piece that is only about 15 minutes, and with that one I de-interlaced and up-res'ed it to 720p30 using Compressor again. I'd say it doesn't look any better than the 480p30 uncompressed video, however there's one big difference. I did some very intensive effects/CGI work on this one to add virtual lighting and atmospheric effects and wanted to do that at a higher resolution than 480p. Decided it wouldn't have been worth it to go up to 1080p, and it would have involved longer render times and bigger files.

IMO, up-res'ing won't help your digital zoom effects, if that's your main concern. The video will already be up-res'ed by anyone watching it on a modern computer or TV. I suppose you could make the argument that you'll have more control on exactly how that is done if you up-res it yourself though. For me, I concentrated on getting the highest native quality from the original, raw video which meant de-interlacing and converting to un-compressed before doing any editing.

Just to clarify... although I went up to 720p, I kept the original 4:3 aspect ratio, so the finished video has an active area of 960 x 720 with black bars on the sides. That seemed like as far as I'd want to stretch SD video. And since HD uses square pixels, the result is a 50% vertical stretch but only a 33% horizontal stretch of the original.

While I've never used the software, to echo what Boyd said I've never had seen any benefits of increasing the resolution. Occasionally I have to playback low resolution surveillance footage and clients often want to know if it can be enhanced. The software is free so try it out and let us know if it does anything. In my experience any software that delivered results would charge a premium for that ability.

Thanks Boyd. I don't really understand the process you outlined...never have understood interlacing...

To clarify...my editing is complete and each episode of the project has been rendered into one file. It is in 16:9 format...720x576, PAL, and I would not want to be up resing all that far, certainly not to 4K. Files sizes vary from about 3GB to 8GB

Yes, the only parts I'm really concerned about are those where I have used digital zoom. Most of these are various angles of me as a talking head. Not all of the image is bothering, only the outline of the head and shoulders which can be jaggedy. The centre of the face and the body of clothing seem fine...well OK. These sections of each episode were all shot in front of a green screen (probably doesn't help) so that has been extracted and replaced...

I did a google search and found a comment by a member here, Mike Kub...? (forgotten how to spell his name - sorry), who pointed to a Boris BCC Continuum unit, Boris Uprez, which does up resing...

Renton

Without seeing the project but if its that simple as you speaking in front of the camera, why not reshoot it? If you waited 15 years what's an month work to re shoot the talking head part? You'll know exactly what to say...

To clarify...my editing is complete and each episode of the project has been rendered into one file.

Well then, to keep it simple, I don't think you would see any benefit at all from messing with that video. More likely, anything you'd do at this point will actually degrade the quality. As I mentioned, modern TV's are going to automatically up-res it to their own native resolution. I would just settle for that.

To put all the details aside, the point I was trying to make was that (IMO) the way to improve old SD footage is to put it in the best possible format BEFORE you edit.

I have come across:
Video Enhancer: resize video with Super Resolution method from SD to HD (http://www.infognition.com/videoenhancer/)
which purports to to be able to do this, but I was wondering if anyone has any experience in this process, or knows of other programs to do this, and perhaps could guide me in regards to it.


I recall testing the AVISynth port of Infognition 'Super Resolution' filter years back and wasn't overly impressed.

For upscaling HD (1080p) to UHD (2160p) I prefer to use AVISynth routines that utilize NNEDI3 as the resizer - in particular a function ('SuperRes') that is available in a plugin that runs HLSL pixel shaders. Full explanation of how it works:

https://forum.doom9.org/showthread.php?t=172698

With relatively clean, good quality HD sources it yields very nice results. I haven't tested it with SD sources myself, but with interlaced material you do need to deinterlace first (single or double rate) - I'm pretty sure that applies to the Infognition 'Super Resolution' filter also.

I would suggest looking at the 'Super Scale' feature in DaVinci Resolve. In Resolve 16 it it requires the licensed (paid) Studio version. However, in Resolve 15 (which is still available for download) it was available in the free version also. The improved 'Super Scale' algorithm in Resolve 16 does recover a bit more fine detail but the version in Resolve 15 is still very good.

What's nice is that it sharpens and denoises in the same process and the strength of sharpening and noise reduction can be set independently from low to high (default Medium). Moreover, it can be applied to directly interlaced sources when configured for 'field processing' - so doesn't require deinterlace (which is only available in the Studio version anyway). Also you don't need to put clips on the project timeline - it can be applied purely for file conversion using Resolve's Media Management >Transcode function. It is very computationally intensive though and so quite slow.

Greetings all and a happy new year...

I have finally finished the editing of a video project I've been working on for...something like 15 years.

Man alive if I had invested 15 years in a project I would hire a Teranex for a day or two and do the job to the highest degree possible. Especially if the early footage was interlaced footage. Today Terenex conversion is one of the industry standards. When Teranex first came out it was like half a million dollars. Today you can buy one for $1,695 so even if you had to buy one you can avail yourself of a very high quality of conversion. A small investment for 15 years of work. If not required afterwards sell it.

Chris Young

https://www.bhphotovideo.com/c/search/BI/2855/KBID/3801

https://www.blackmagicdesign.com/au/products/teranex

The AVISynth/VapourSynth deinterlace filter, QTGMC, also produces exceptionally good results, for free:

QTGMC - Avisynth wiki (http://avisynth.nl/index.php/QTGMC)

Man alive if I had invested 15 years in a project I would hire a Teranex for a day or two and do the job to the highest degree possible. Especially if the early footage was interlaced footage. Today Terenex conversion is one of the industry standards. When Teranex first came out it was like half a million dollars. Today you can buy one for $1,695 so even if you had to buy one you can avail yourself of a very high quality of conversion. A small investment for 15 years of work. If not required afterwards sell it.

Chris Young

https://www.bhphotovideo.com/c/search/BI/2855/KBID/3801 (https://www.bhphotovideo.com/c/search/BI/2855/KBID/3801/BI/2855/KBID/3801)

https://www.blackmagicdesign.com/au/products/teranex

It is actually longer than 15 years! Until I started on this I had never held a video camera. I spent eight years acquiring gear and learning how to use it, and learning filming before I even started on the actual project. I've been editing for about 5 years. This would have cost mega bucks if done commercially! I've done it on the smell of an oil rag...as they say.
Your suggestion has given me some ideas which I will pursue...thanks. You're right. Pity to stumble at the last hurdle...
Renton

Incidentally, came across this thread on the Creative Cow > Vegas Pro forum that addresses 'Smart Upscaling' of SD:

https://forums.creativecow.net/thread/24/1024203

One post there points to a set of tools (Happy Otter Scripts) for Vegas that utilizes open source filters including AVISynth and VirtualDub plugins:

https://tools4vegas.com/

I've just had a talk to a friend who is in the video hire and eventing scene who has a Teranex, but he's suggesting it is not actually uprezing that I need to do but just removal of noise on the clips where I've done digital zoom. I've never been this way before but he suggested it could be as simple as applying a noise filter to the particular clips.

Any thoughts re doing something like this in Vegas Pro? Could this be the solution?

Seeing I'm going to deliver on DVD (I think), apparently uprezing is superfluous...

Renton

Seeing I'm going to deliver on DVD (I think), apparently uprezing is superfluous...

That's a "little detail" you probably should have mentioned earlier. There is clearly no point in up-res'ing and then down-res'ing back to standard definition. That will definitely just degrade the video.

... he's suggesting it is not actually uprezing that I need to do but just removal of noise on the clips where I've done digital zoom.

Any thoughts re doing something like this in Vegas Pro? Could this be the solution?

Seeing I'm going to deliver on DVD (I think), apparently uprezing is superfluous...

Renton

This is a different story. If DVD is the final destination and it is going to be PAL, or NTSC for that matter then yes absolutely no point in up-rezzing the video.

I've spent the last 12 years working on military history docos where using tons of old historic footage is the norm. A lot of this is from film in every conceivable frame rate and aspect ratio you can image. A lot of these films have been converted to all sorts of codecs and wrappers. The last series had around 3000 hours of footage to cull through. A lot of the chosen material was noisy as you would expect. All of this old 4:3 material had to be standards converted and panned and scanned to 16:9.

When it came to the noise reduction on all this archive material we tried hardware and software NR. The best results by far came from using 'Neat Video' for the NR because of its customizability. You can do a simple profile sample which works quite well for certain material but some material really needs scene sections selected to build a custom profile to get the best results. Neat Video is my recommendation for most video that requires NR. Neat also retained detail better than anything else we tested. It is GPU assisted and is available for many NLEs. Even with with a late model high performance platform you can expect slowish renders because the whole Neat Video NR process is computationally very demanding.

Chris Young

https://www.neatvideo.com/

Basic Workflow | Neat Video 5 - YouTube

I think if you were working with high quality HD/4K sources and zooming in 200% or more there could be a case for applying these detail preserving, super-resolution techniques. But for SD, I agree with the others, I don't think you'd gain any benefit and it would likely make matters worse.

As for denoising. Again, personally I prefer using AVISynth/VapourSynth techniques that best target the type and degree of noise that is present. Fine, grainy noise is relatively easy to treat. Blotchy colour-noise is rather more difficult without obliterating detail and introducing more artifacts in the process. Bear in mind also that the human eye is an amazing integrator and can often perceive detail amid noise in moving images that may not be revealed in a single frame especially when viewed at a distance. So it comes down to a compromise of what is the most acceptable 'psychovisual' outcome on the target viewing platform. Maybe consider also if the best outcome is not zooming in on your talking head shots or limiting it to a degree that retains acceptable IQ. In any event, it is best to denoise before any digital zooming/enlargement.

Haven't used Neat Video myself but I know many extol it's virtues.

Thanks Christopher.

Just had a look at Neat Video and I'm wondering if I've used the correct terms. The 'noise' as shown on their samples of removal, is not what I'm primarily wanting to remove, although there may be some of that there. Rather it is blockiness/pixelation on the edges/outline of my clothes and head which I would like if possible to smooth out.

Renton

Could you post a screen shot to confirm what you are referring to specifically - is it 'pixelation', which is normal when you zoom-in/enlarge a digital image enough to see the individual pixel blocks or is it 'macroblocking' (a compression artifact, sometimes referred to as 'block noise') when objects or areas in a video image appear to be made up of irregular sets of small squares, rather than proper detail and smooth edges.

It is pixelation, not block artefacts...

I have had one interesting situating come about (but this is different to above) where I have rendered out a clip in Vegas to a new track, then on rendering the whole thing out to an .avi file, some artefacts have appeared over a couple of frames in the clip where they never showed before...

I've just has a look at one of the episodes again and think that I will simply leave them as they are.

Thanks for all the ideas offered, even though I am now not going to try anything further. Appreciated.

Now it is on to working out the DVD arrangement for all the files...

It is pixelation, not block artefacts...

OK, so it's only when you zoom-in then? In that case, since you are not 'up-resing', the standard 'treatment' would be to blur (Gaussian) and then contra-sharpen to see what detail you can pull back. A trade off, as always.

I've just has a look at one of the episodes again and think that I will simply leave them as they are.

Sometimes it's best leave well enough alone.

I know what he’s describing. I once borrowed a pocket camera. I didn’t realize until later that digital zoom was enabled. Everything looked fine while I was shooting but once I imported it on my computer I could see all the zoomed in portions looked awful. As the op said it’s blocky and pixelated, there’s very little you can do to fix it other than to reshoot it.

I won't be reshooting... would mean virtually the whole project would need to be redone from scratch... which aen't gonnna happen :-)