
- Time
- Post link
@ Laserdisc Master
Wow, those look really good. Would be interesting to see a video. I've been working on super resolution v4. Screenshots coming up...
@ Laserdisc Master
Wow, those look really good. Would be interesting to see a video. I've been working on super resolution v4. Screenshots coming up...
Here are the first screenshots for super resolution v4. I'm rather proud of the result. It has enormous detail, but little noise or artifacts. I think it achieves a natural high definition look. What do you think?
These screenshots are all from the Tantive IV scene. I think you all would agree that was a pretty noisy scene. Here's what I've been able to do with it:
http://screenshotcomparison.com/comparison/122567
http://screenshotcomparison.com/comparison/122568
http://screenshotcomparison.com/comparison/122569
@DrDre
Woah, that is stunning.
You can clearly see the blood vessels in Leias eyes in the one screenshot. I am astounded, incredible work!
@ Zyrother
Thanks! I'm pretty happy with it. But part of the credit goes to Laserdisc Master, who continues to push the envelope with his incredible algorithms.
That is quite an increase in fine detail and clarity. It's always surprising how much detail and good color there is in the 2006 DVD set. The trick is bringing it out without introducing (or exaggerating) other problems in the process. Well done. :)
The Star Wars trilogy. There can be only one.
Here are three more screenshot comparisons for super resolution v4:
http://screenshotcomparison.com/comparison/122572
@ Dunedain
Thanks! I'm also quite surprised by the vibrant colours that come to the surface after the detail recovery and noise reduction, especially since I did nothing the explicitely alter the color palet of the source.
Kind of off topic, so I apologize in advance. But which do you think has color that is closer to the original elements? The Blu Ray or the GOUT?
With the Blu Ray, they seem to over-saturate the colors so they "pop" more. Whereas the GOUT seemed to capture the natural color palette better, even though the transfer was very substandard.
Digital video to me seems over saturated and makes it hyper real, whereas film has a soft, and very natural color. Plus, I love film grain. :)
Zyrother said:
Kind of off topic, so I apologize in advance. But which do you think has color that is closer to the original elements? The Blu Ray or the GOUT?
With the Blu Ray, they seem to over-saturate the colors so they "pop" more. Whereas the GOUT seemed to capture the natural color palette better, even though the transfer was very substandard.
Digital video to me seems over saturated and makes it hyper real, whereas film has a soft, and very natural color. Plus, I love film grain. :)
Neither are really good sources to the original color, but the GOUT is probably closer. Someone here has got an LPP print lying around somewhere, which would probably be closest to the original color. Stay tuned! ;)
Here you are a small test clip: https://www.sendspace.com/file/ixquyz
Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com
Impressive... most impressive.
Latest comparison - SRV4 Vs MU-something (^^,)
http://screenshotcomparison.com/comparison/122294
happycube said:
Impressive... most impressive.
Think to this technique used with your laserdisc capture workflow...
Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com
Holy cow, the noise is nearly gone, yet so much detail is still remaining.
Magic, cant be anything else. :P
@ Laserdisc Master
Could you show what your method would do with the soldiers face in the following screenshot I posted before for super resolution v4? I'm curious how it would handle the facial texture, and the lack of smooth shapes.
DrDre said:
@ Laserdisc Master
Could you show what your method would do with the soldiers face in the following screenshot I posted before for super resolution v4? I'm curious how it would handle the facial texture, and the lack of smooth shapes.
Here you are:
http://screenshotcomparison.com/comparison/122296
Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com
@ Laserdisc Man
You accidentally compared your upscale to the Avisynth Spline64Resize, not to the super resolution v4. Here the correct comparison:
http://screenshotcomparison.com/comparison/122514
As before your upscale is cleaner with less noise, but mine contains more texture and detail. Interesting comparison!
Here's a first video sample of the super resolution v4. It's part of the Tantive IV scene.
Low res source:
https://drive.google.com/file/d/0B8_LYKyZDiajVVlQQlE2OTVITVE/view?usp=sharing
Super resolution v4:
https://drive.google.com/file/d/0B8_LYKyZDiajYkhnUEg4X2pGb2s/view?usp=sharing
Please download for full quality.
DrDre said:
@ Laserdisc Man
You accidentally compared your upscale to the Avisynth Spline64Resize, not to the super resolution v4. Here the correct comparison:
http://screenshotcomparison.com/comparison/122307
As before your upscale is cleaner with less noise, but mine contains more texture and detail. Interesting comparison!
Thanks for getting the correct comparison. Thought I was going crazy before. :P
Yours does have a bit more detail. I guess it's just a question of where the tradeoff should be. The chinstrap looks so much nicer in LM's, for instance, as that double halo artifact on the chin is pretty bad. The eyes, though have a hint more detail in yours.
@ towne32
Yes, the enhanced halo is unfortunately a side effect of the super resolution. I wrote my own super resolution algorithm in Avisynth based on the Infognition plugin. This algorithm dramatically enhances the detail compared to the standard super resolution. The halo is present in the source in most frames, so sadly the algorithm interprets it as cross frame detail, so it get's enhanced along with the rest. I implemented an advanced noise filter to counter the effects of noise enhancement. The halo however can't be deminished without compromising the detail and crispness of the upscale. To me it's an acceptable cost, but perhaps others will prefer the smoother alternative.
DrDre said:
@ Laserdisc Man
You accidentally compared your upscale to the Avisynth Spline64Resize, not to the super resolution v4.
My fault, sorry!!!
Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com
A new, "dirtier" version... this time I've left behind any noise reduction, and refined my SuperResolution filter, involving also the use of other kind of resizers...
SuperResolution V4 Vs MagicUp SR2:
http://screenshotcomparison.com/comparison/122417
- hope this time it's the right image! (^^,) -
Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com
_,,,^..^,,,_ said:
Think to this technique used with your laserdisc capture workflow...
Oh I have been... :) (If this can make the GOUT look this nice, imagine how JSC might look!)
I'll probably look at http://mentat.za.net/supreme/ this weekend, which is BSD-licensed. There are some single-frame and neural network techniques that sound interesting as well.
happycube said:
_,,,^..^,,,_ said:
Think to this technique used with your laserdisc capture workflow...
Oh I have been... :) (If this can make the GOUT look this nice, imagine how JSC might look!)
Yep! Also, consider the various PAL versions... they are good, but because PAL players are not high-end, your invention could squeeze a lot more quality from them!
I'll probably look at http://mentat.za.net/supreme/ this weekend, which is BSD-licensed. There are some single-frame and neural network techniques that sound interesting as well.
I was aware of this, but not able to use it... if this technique could mantain the promises as seen in the pictures, it could be the real winning one!
Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com
_,,,^..^,,,_ said:
Yep! Also, consider the various PAL versions... they are good, but because PAL players are not high-end, your invention could squeeze a lot more quality from them!
I'm in the US so PAL isn't really available to me, but I do have some good PAL RF samples and every so often I make some progress on PAL stuff. I think I'll look at it more this next week(end).
I'll probably look at http://mentat.za.net/supreme/ this weekend, which is BSD-licensed. There are some single-frame and neural network techniques that sound interesting as well.
I was aware of this, but not able to use it... if this technique could mantain the promises as seen in the pictures, it could be the real winning one!
Perhaps... I pretty much only use Linux, so can any of the scripts you or Dre use be ported to Vapoursynth?
happycube said:
I'm in the US so PAL isn't really available to me, but I do have some good PAL RF samples and every so often I make some progress on PAL stuff. I think I'll look at it more this next week(end).
That's great!
Perhaps... I pretty much only use Linux, so can any of the scripts you or Dre use be ported to Vapoursynth?
Don't think so... maybe you could port that technique to AviSynth... (-^,)
Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com