Oke, super resolution v9 will have reduced aliasing. I have found an antialiasing filter that reduces the aliasing compared to the raw GOUT and leaves the detail intact. The differences will be subtle, but noticeable.
The AA filters I've tested do better, but the loss in terms of detail is still apparent, so some progress, but not so much that I would use any of them at this point. I think I will now first focus on releasing a version of this that goes for maximum detail recovery, with a mild grain reduction, and leave the antialiasing for the next iteration. I never considered doing multiple iterations, but I've discovered that once you're down this path, forever will it dominate your destiny... ;-)
The laserdisc captures I've seen so far have a significantly lower resolution than the GOUT. However, I've seen some pretty promising developments on this forum with respect to laserdisc captures, so if one approaches the GOUT in terms of detail, it is certainly worth a try.
Here's the updated 10 min sample for the debugged super resolution v8:
https://drive.google.com/file/d/0B8_LYKyZDiajakI5STdaTEdyRVU/view?usp=sharing
There's more detail than before. Sadly, aliasing remains an issue.
The scenes with R2 in the canyon are very good frames to show super resolution detail reconstruction. Take this frame fo example:
Yes, unfortunatelly antialiasing and super resolution are not best friends. I did make some alterations in the SR script I use, which seems to reduce the aliasing to a certain degree. Maybe your antialiasing will work better on the sample I put up. I would be happy to share my SR script if that helps.
Here's the same frame compared to Harmy's Despecialized Edition 2.5. I hope Harmy won't mind.
I'm posting a single comparison with Team Blu, to highlight the differences between the two approaches. The methods chosen to enhance the film leads to a very different end result. In my humble opinion there are positives and negatives to be found in both methods, which is why they can co-exist as two different visions on what the GOUT can be.
http://screenshotcomparison.com/comparison/127122
Anyone interested in other comparisons, please send me a PM or post on this thread,
Yep, that's the one. Here is a long list of screenshot comparisons for the debugged SRV8:
http://screenshotcomparison.com/comparison/127093
http://screenshotcomparison.com/comparison/127094
http://screenshotcomparison.com/comparison/127095
http://screenshotcomparison.com/comparison/127096
http://screenshotcomparison.com/comparison/127097
http://screenshotcomparison.com/comparison/127098
http://screenshotcomparison.com/comparison/127100
http://screenshotcomparison.com/comparison/127102
http://screenshotcomparison.com/comparison/127103
http://screenshotcomparison.com/comparison/127105
Update:
Here are the comparisons to the simple upscales for the frames used in the MagicSR comparisons:
http://screenshotcomparison.com/comparison/127115
The video is being processed, and I will put up video samples as soon as possible.
It's debatable, but especially the Tantive IV scenes, and the very grainy scenes on Tatooine benefit from slightly less grain, especially in motion, as could be seen in the last video sample. Also, the colors are brought out more.
Andrea, I can't thank you enough. There actually was a bug in the SRV8 script, causing some of the detail reconstruction to be undone. So, the full potential of SRV8 has not yet been seen. It sadly also means I have to process the movie again... :-(
Here are new comparisons to MagicSR for the previous frames:
http://screenshotcomparison.com/comparison/127050
http://screenshotcomparison.com/comparison/127051
http://screenshotcomparison.com/comparison/127052
There are actually tiny new details to be found here and there...
Update:
Removed another bug, so there's a little more detail.
I don't use any blurring. Super resolution V8 is a combination of a super resolution script and non-linear filter I wrote, combined with a anti-ringing filter.
The frames I used for the first three are: 17119, 22100, and 23755.
This screenshot comparison shows very clearly that super resolution has some clear advantages over sharpening. In this case I used the sharpen filter in Virtualdub for the comparison. From this example is it very clear that compared to sharpening, super resolution has more detail, while having less noise enhancement, less ringing, and less aliasing.
Avisynth Spline64Resize vs Avisynth Spline64Resize + sharpen
http://screenshotcomparison.com/comparison/126997
Avisynth Spline64Resize vs super resolution (raw)
http://screenshotcomparison.com/comparison/126998
Avisynth Spline64Resize + sharpen vs super resolution (raw)
http://screenshotcomparison.com/comparison/126999
So, unlike some critics stated in the past, super resolution does indeed enhance detail, while simultaneously reducing noise, ringing, and aliasing compared to standard enhancement techniques.
I hope to get another video sample online this weekend. In the mean time here are a couple of screenshot comparisons for the sample. Notice the detail recovery for R2D2 in the first screenshot.
http://screenshotcomparison.com/comparison/126843
http://screenshotcomparison.com/comparison/126844
http://screenshotcomparison.com/comparison/126847
Update:
Here's a screenshot for one of the scenes I'm working on now:
http://screenshotcomparison.com/comparison/126862
This is a standout example of super resolution detail reconstruction, especially noticeable on C3PO.
Great cleanup! It looks so different from the digital mess, the bluray release of Star Wars (and not just the colors). Makes me feel all warm inside... ;-)
2 TV's Frink
You're absolutely right! Let's use 2 instead. :-P
@ Zyrother
This is one of the challenges with this type of upscale. Some of the scenes are so grainy, the amount of detail enhancement is matched by an equal amount of grain amplification. The standard filters weren't quite up to par, so I wrote a non-linear filter that keeps as much of the detail, while removing the enhanced noise, and replacing it with a reduced GOUT grain.
@ zee944
Due to the filtering a tiny amount of detail is lost. This frame always had a relatively small amount of visible grain due to the dark background. For light backgrounds the difference is pretty obvious though. Here's an example:
@ zee944
The differences between the simple upscale and raw super resolution are still pretty easy to spot (at least to my eye).
Update:
Here's the comparison between SR raw and SRV8:
http://screenshotcomparison.com/comparison/126617
update 2:
I now see it's not the same frame. It's difficult to see on my mobile. I used a different frame for the Team Blu comparisons than the ones I posted earlier.
@ zee944
The earlier comparison included a combination of super resolution and non-linear filtering (super resolution v8). This is the raw super resolution result that enhances detail, but also enhances noise (super resolution v3).
Here's the same comparison between sharpening, deblurring, and super resolution for the Obi-Wan frame I posted a while ago.
Simple upscale vs simple upscale + sharpen
http://screenshotcomparison.com/comparison/126592
Simple upscale vs simple upscale + deblur
http://screenshotcomparison.com/comparison/126594
Simple upscale + sharpen vs simple upscale + deblur
http://screenshotcomparison.com/comparison/126599
Simple upscale vs raw super resolution
http://screenshotcomparison.com/comparison/126602
Simple upscale + deblur vs raw super resolution
Thanks, I will also add it to fanres, later today. :-)
To clarify the difference between sharpening, deblurring, and super resolution, I will use another example programmed in MATLAB.
When an image (or video frame) is downscaled two things happen: the image is blurred and compressed. Blurring causes the information content to be distributed differently among the pixels, resulting in loss of detail, but in principle no information is lost. Compression results in loss of information. We can subsequently upscale the downscaled image back to its original size through interpolation. When we compare the upscale to the original, we obviously get a blurred image with less detail:
http://screenshotcomparison.com/comparison/126569
Sharpening is the same as applying an unsharp mask filter. An unsharp mask filter enhances the high frequency content in the image. In other words sharpening results in edge enhancement. No lost details are recovered, the existing details are simply made more visible:
http://screenshotcomparison.com/comparison/126570
A side effect of sharpening is ringing. Although sharpening enhances detail, it does not necessarily result in a better representation of the original high res image. It simply represents a subjective reimagining of the low resolution image.
Deblurring is the process of undoing the effects of blurring. Provided a reasonable estimate of the blur function is known, many details can be recovered:
http://screenshotcomparison.com/comparison/126571
The following comparison shows the difference between sharpening and deblurring. The deblurred image obviously has more detail, and is a better representation of the high res image. The deblurred image also has less artifacts, such as reduced ringing:
http://screenshotcomparison.com/comparison/126572
However, the loss of detail due to compression cannot be undone by deblurring. If we have multiple images with subpixel shifts the images can be super resolved, and part of the detail lost due to compression can be recovered. The objects in the images are registered, aligned, averaged, and deblurred:
http://screenshotcomparison.com/comparison/126585
The super resolved image has more (accurate) detail, less artifacts, such as reduced ringing, and is an even better representation of the high res image, when compared to the deblurred single image:
http://screenshotcomparison.com/comparison/126586
Summarizing, sharpening enhances existing details, deblurring reconstructs details lost as a result of blurring, super resolution reconstructs details lost as a result of blurring and compression. Any of these methods results in artifacts, the foremost being ringing. However, sharpening generally has the most artifacts, while super resolution has the least.
The sucess of super resolution in practise, hinges on accurate data fusion, and the accuracy of the estimated point spread function (blur function). Over the last decades increasingly accurate approximate solutions have been defined for both these problems, resulting in accurate and patented applications in image/video processing and forensics.
For more information on the limits of super resolution, I refer you to this scientific paper:
http://homepage.tudelft.nl/e3q6n/publications/2005/SPIE2005_SanJose_TPLVKS.pdf
@ Laserdisc Master
Sorry I forgot to reply to your post. Now I was being rude. I will ignore myself. :-P
The example I used is highly idealized, but the SR algorithm I use, uses the same principle. This is the registration step, where similar objects in different frames are identified and aligned.
@ TV's Frink
I think you forgot my bread...