there is no difference for me

[mario]

I am tryig to uprez my sample file from SD to HD. First step: I deinterlace my file in Sony Vegas, then I open VENhancer and I apply sr filter to 1440x960; then I use uprezed file and run it again through sr filter again uprezing to 2880x1920...and these two files - original and final looks almost the same; I know I am doing something wrong; can somebody post simple "template" how to uprez it to hd using Venhancer; step by step using fileter as found on their web site. For me their explanations like" upsize all frames 2 times using good spatial interpolation, in this case Lanczos3 method;  using sub-pixel accurate motion compensation find similar areas in neighbor frames and intelligently merge frames to combine information (and in some cases reduce noise; combine neighbor frames to get more details using motion compensation ---- sounds like hi tech mombojumbo. let produce simple method using oarticular filters. anyone???

[Efenstor]

Sure, there is not MUCH difference, because in fact super-resolution can give only about 10-60% of brand new information (depending on source quality). VideoEnhancer is for you only if you need the "as-good-as-possible" results, to be sure that that's the best you can get. I did comparisons and it's indeed better than any spline interpolation but not that better to make you instantly say "wow".

[Irina]

Mario:
you are quoting an article "What is super resolution" which aims at general explaining how SR algorithm works. It would be a better idea to read a new manual for Video Enhancer:
http://www.thedeemon.com/VideoEnhancer/manual.html
Cheers!

[Zartan]

Hopefully by v.2 the SR will be better. Sometimes it won't remove noise and leaves some artifacts behind. I have used Videoenhancer alot and It's probably better than other resize methods. Try 1.9.2 and compare it to a Lancose3 there's not much difference. We need an Ultra SR setting, Full and Fast just isn't cutting it. 

[Dee Mon]

SR in its current form will not remove noise. Just the opposite: it thinks the input video contains good information and uses it to restore higher resolution video. If there is some noise, it will try to upsize the noise as well making it even more annoying. And since the noise is different in different frames, it degrades overall SR quality. So it's highly recommended to use denoising filters before SR if you deal with noisy video.

As for Ultra SR, this will definitely be added soon, but the speed of this mode may be much worse. The speed is the only reason why there weren't many improvements in SR lately: many people complain it's too slow already. It is possible to get higher quality, but it's impossible to do so without hurting performance.

Probably if we implement SR on GPU we would be able to have both quality and speed, but this transition hasn't even been started yet.

[Efenstor]

Please don't worry that much over speed, implement Ultra SR, I'm sure I'm not the only one who agrees to spend time to get real quality!

[itou]

I also would like an ultra SR option. Quality is why I choose VE (for SR).
I used to up my 720p@60 fps to 1080p@60 fps (as my 1080p mode cam is only 30 fps).
The other strength of VE is that it able to take dshow file directly (HDV mov file). But the output xvid is not "conform". I don't know why but xhen I convert a HD mov file mp4 to xvid (throught vdub using avisynth directshowsource or mediacoder), the file is readable into an old video editing software (stereo movie maker) and not the one convert by VE (1.9.2 or earlier), even after changing FOURCC "XVID" to "xvid". Sometime it read but with very weird blocky color square artefact as seems it missed some I frames ...
I've tried on other computer with different xvid codec but have the same issues.
So my solution is now :
- mp4 mov 720p --> VE + vdub filters --> lossless lagarith 1080p (huge size !!) --> mediacoder --> xvid --> stereo movie maker
I will take a look to frame server to avoid lossless stage.

[Dee Mon]

Have you tried DivX instead of XviD? Maybe SMMaker will read it, so you don't have to make extra steps...

I guess one should expect Ultra SR this autumn.

[Manoman455]

Please don't worry that much over speed, implement Ultra SR, I'm sure I'm not the only one who agrees to spend time to get real quality!

Yes I would love an ultra SR mode. Speed isn't a problem for me I use so many filters it takes the whole day to process anyways, I can wait.

[Sunny666]

I use SR and then MSU Deblocking 2.2. Picture is really much more stable and sharp. It takes about 100 minutes to transcode 40 minutes of 624x352 video on quad-core Phenom II x4@3400. Can anyone confirm that new Intel Core i7 processors perform better and how much?

[dedica]

Please don't worry that much over speed, implement Ultra SR, I'm sure I'm not the only one who agrees to spend time to get real quality!

Yes I would love an ultra SR mode. Speed isn't a problem for me I use so many filters it takes the whole day to process anyways, I can wait.

I've got to chime in for the reverse side here.

The super resolution filter is already the strongest on the market. If you're using HQ mode this already does incredible things. (As long as you're not feeding it block-compressed or noisy video, which no amount of super resolution is going to tackle anyway.)

It's hard to discount speed here. This complex process is already brokenly slow on a cpu. There are a ton of files that I haven't even processed because of time. Even short vids can take a whole night. Imagine being able to tackle your longer videos without any problems. GPU integration would do much more right now than a tweak to the sr.

I'm wondering if the OP may be using blocky or noisy video without realizing it, and without applying deblocking or denoising filters first. (720 x 480 vid isn't necessarily always block/noise free, especially if it was crappily upsized from smaller video before you recieved it.) Either way, it sounds like the author is saying that another level of super res would take (A LOT) longer than it already does. Plus, the sr is already incredible if you feed it correct video.

[Manoman455]

It takes me about 4 hours for a half hour video. I really don't mind waiting even 24 hours for better quality. besides its just adding another mode not replacing any, so people with slow computers (no offense) can still used the speed and high quality mode and leave the ultra-high quality alone.

[dedica]

However, not everyone is eager to sacrifice extreme speed for a bit more quality.

Being able to have both the current quality (which is already insanely high) and speed
has something for both tastes.

Also not sure what kind of processing that 30 minute video is going through, but a
C2D E8400 at 3.0 GHz (which certainly isn't slow) can take hours for 10 minutes of
decent-quality conversion. (a chain of two SR filters both in HQ mode, plus maybe
a couple of other filters for noise, dot crawl, blur or other picture issues.)

A major upgrade in speed with no downside vs a small, time expensive increase in
quality. The benefit of your gaming card being able to speed up something as scientific as
this instead of just games.  I know I'd rather be able to process 10 long files in a day than
get a 10-20% increase in 1 file that takes all day, plus put my video card's power into
video apps.

Being able to use this day and night and still have time to use your computer's power for
gaming and other things would be another giant plus.

[Dee Mon]

>a chain of two SR filters both in HQ mode

This makes sense only if you upsize more than 2x. Is it the case?

[Manoman455]

I really hope that the new version is completed soon, this version has bugs that takes me a while to work around. Not major bugs....again its with the codecs!!!