Thanks, I'm working on upscaling the NTSC GOUT. It's looking pretty good. Star Wars is being upscaled as we speak, and Empire is next in line.
You could try that, but I would say that a 1080p video from a true 4K source will always be better than one from a super resolution 4K upscale. Although you can reconstruct a good part of the detail with super resolution, some information is always lost in the downscale that can never be retrieved. The 4K upscale from 1080p would contain more detail than the 1080p original, though.
@lpd You can apply the algorithm to the 1080p original. In that case you would get a 4K ultra high definition video. I would expect it would be about as accurate to the true 4K film (if it were ever released) as the 1080p upscale is to the true original 1080p transfer, provided the 1080p transfer is of good enough quality.
I guess once the powers that be decide to give us the blu ray's of the theatrical editions, this is the next project in line: a 4K upscale for those of us with a UHD TV.
The great thing is that anyone can try it for themselves, and it's done in a few hours. The infognition software has a free demo for 30 days with full functionality.
Here is another screenshot comparison. From these examples it is very clear the super resolution is able to reconstruct texture in the skin and clothes, hair, eye lashes and other details. To say that it is only acts as a noise filter is shortchanging many years of scientific research, in my opinion.
1080p original vs super resolution:
http://screenshotcomparison.com/comparison/118506
Precise bicubic (A = -0.75) vs super resolution
http://screenshotcomparison.com/comparison/118507
1080p original vs precise bicubic
1080p original vs super resolution:
http://screenshotcomparison.com/comparison/118497
Precise bicubic (A = -0.75) vs super resolution:
http://screenshotcomparison.com/comparison/118500
1080p original vs precise bicubic (A = -0.75)
Here is a test of the super resolution upscaling on the 2011 Star Wars blu ray. The original sample is 1080p and has been resized to 540p at 8000kps. Subsequently the lower resolution sample has been upscaled to 1080p with the super resolution algorithm.
Update: there is an audio synching problem with the video files. I will fix them and put them back up.
Below are some screenshot comparisions between the original 1080p, the super resolution upscale, and the precise bicubic upscale.
Here is a sample of the PAL GOUT upscale SR-v2.
Update: Clips removed to make space for new clips.
Today I will also start upscaling the NTSC version of the GOUT. As You_too said, it apparently has more detail than the PAL version, and it's therefore better upscaling material. It will also provide for a fair comparison with Team Blu's impressive results.
Of course, if you can send me a clip I'll run it through the algorithm.
For those interested in finding out why multi-frame super resolution is possible, the first chapter of this phd thesis provides some explanations:
http://www.robots.ox.ac.uk/~vgg/publications/papers/pickup08.pdf
@dickalan when I click on your link it just goes to collider.com
I will post a few clips of the super resolution v2 soon.
I'm in the process of downscaling a clip from the Star Wars blu ray, such that we can compare the upscaling results objectively by calculating PSNR's for individual frames. I will post the original HD, the downscaled, and upscaled clips here.
By going to RGB mode of the super resolution software (which is slower, but more precise) you get rid of the edge artifacts, as can be seen in the following comparison:
http://screenshotcomparison.com/comparison/118312
You also get slightly more detail:
The basic concept for super resolution used for fluorescence microscopy and digital sensors is the same, namely combining multiple low resolution images to obtain a high resolution image. In the first case it is called optical super resolution. In the second case it is called geometrical super resolution. Since a video is nothing more than a sequence of very similar images, the concept has been successfully applied to video as well, for example in forensics. Wikipedia actually explains both pretty well:
I think one 2014 Nobel prize winner would disagree:
http://www.nobelprize.org/nobel_prizes/chemistry/laureates/2014/advanced-chemistryprize2014.pdf
The BasiClean+MagicUp has less grain, but also considerably less detail, visible in Obi-Wan's beard and Luke's hair, but especially in the string in the left corner of the frame on Obi-Wan's hood.
Another comparison (two video samples) between precise bicubic (A = -0.75) and the super resolution.
Update: Clips removed to make space for new clips.
Two more comparisons for precise bicubic (A = -0.75) vs super resolution. In this case both sources are uncompressed. In both cases details are more clear in the super resolution version. The grain also seems to be somewhat more visible though, and a number of edge artifacts are clearly visible in the second screenshot, which is of course very undesirable.
Original GOUT vs super resolution
I added a fair comparison: precise bicubic (A = -0.75) vs super resolution
The differences are subtle, but you can see more detail in the eyes, eye lashes, hair in the super resolution one. You could try sharpening the bicubic, but this would also increase the grain/noise.
I will convert a scene from the 2004 dvds to compare with the 2011 blurays.
Which software do you use to get the individual frames? The raw video is a bit too large to put on google drive.
...and a first sample for The Empire Strikes Back:
Standard definition:
https://drive.google.com/file/d/0B8_LYKyZDiajY194Wm9wblFPSFE/view?usp=sharing
Upscaled:
https://drive.google.com/file/d/0B8_LYKyZDiajeFA0VkZuaENiY1U/view?usp=sharing
Here's another sample:
Update: Clips removed to make space for new clips.