AI Video Enhance
Out of context: Reply #5
- Started
- Last post
- 7 Responses
- monNom0
I don't know how this software works specifically, but there is a technique called 'super-resolution' (OpenCV has an implementation), that takes multiple frames with slight movement between them and builds up a higher resolution image from them. It sort of solves what the underlying image must be based on the variations in pixel values for a given feature across multiple frames. So the whole CSI "enhance" does exist to some degree if you have multiple video frames.
If you then morphed that high resolution image using optical flow from the original video, you could probably get a decent result. It would probably need to be a relatively static scene to work.
Neither of these are AI, but there might be some GAN magic involved in blending edges and inpainting in fast changing scenes. I don't really understand how that stuff works so just a wild guess.
- I was thinking if there are crisper frames in focus where there is no movement and it takes that info & applies it to the blurier parts then it could work.webazoot