cyvideo
Veteran
Not dissimilar to what Sony have been doing with their lower end Clear Image Zoom AI that uses an automated texture synthesis process utilizing a massive database of stored images and textures.
https://www.bhphotovideo.com/explor...he-most-amazing-shooting-mode-you-never-heard
Another approach being used:
http://webdav.tuebingen.mpg.de/pixel/enhancenet/
Check out the following if you need to pull some higher rez images from video grabs.
https://deep-image.ai/
I've been pretty amazed, not that that's hard, by what I have uploaded and then up converted using their AI & machine learning algorithms. Its based around the technology that JCS referred to coming from
https://arxiv.org/abs
"Accelerating the Super-Resolution Convolutional Neural Network"
https://arxiv.org/abs/1608.00367
and
"Compression Artifacts Removal Using Convolutional Neural Networks"
https://arxiv.org/abs/1605.00366
You get five upload/up converts/downloads for free from Deep Image AI to see if you think the process is worth it.
This Topaz Gigapixel AI for video interests me as on some historical doco material I've been using their Topaz Denoise AI (similar tech) on stills and even video where I have exported a stills sequence from a clip and then using Photoshop doing a batch process of the stills using Denoise AI and then re-importing the stills as an image sequence for rendering out as a restored clip. Topaz's Denoise AI on old archive footage has Neat Video well beaten.
There again just looking at re-scaling upwards having used Resolve's Super Scale feature I have been pretty surprised at how good a job that does for most cases when I require an upscale when there isn't much budget available.
https://www.premiumbeat.com/blog/resolve-15-super-scale-feature/
Tech is changing so fast with all these developments that sometimes it makes me wonder if we really need to chase 6/8K cameras when you can dial up what you want in software on decent performance platforms. So much easier to work with smaller dimension and sized files and then output to your resolution of choice. Interesting times indeed!
Chris Young
https://www.bhphotovideo.com/explor...he-most-amazing-shooting-mode-you-never-heard
Another approach being used:
http://webdav.tuebingen.mpg.de/pixel/enhancenet/
Check out the following if you need to pull some higher rez images from video grabs.
https://deep-image.ai/
I've been pretty amazed, not that that's hard, by what I have uploaded and then up converted using their AI & machine learning algorithms. Its based around the technology that JCS referred to coming from
https://arxiv.org/abs
"Accelerating the Super-Resolution Convolutional Neural Network"
https://arxiv.org/abs/1608.00367
and
"Compression Artifacts Removal Using Convolutional Neural Networks"
https://arxiv.org/abs/1605.00366
You get five upload/up converts/downloads for free from Deep Image AI to see if you think the process is worth it.
This Topaz Gigapixel AI for video interests me as on some historical doco material I've been using their Topaz Denoise AI (similar tech) on stills and even video where I have exported a stills sequence from a clip and then using Photoshop doing a batch process of the stills using Denoise AI and then re-importing the stills as an image sequence for rendering out as a restored clip. Topaz's Denoise AI on old archive footage has Neat Video well beaten.
There again just looking at re-scaling upwards having used Resolve's Super Scale feature I have been pretty surprised at how good a job that does for most cases when I require an upscale when there isn't much budget available.
https://www.premiumbeat.com/blog/resolve-15-super-scale-feature/
Tech is changing so fast with all these developments that sometimes it makes me wonder if we really need to chase 6/8K cameras when you can dial up what you want in software on decent performance platforms. So much easier to work with smaller dimension and sized files and then output to your resolution of choice. Interesting times indeed!
Chris Young