Popular sides of video processing

[slavanap]

Hello,
I want to know what is actual subjects in video processing that haven't been fully researched and need to be researched. What video filters you want to use, that don't exist yet?
I really need your opinion!

Best regards,
slavanap.

[Mug Funky]

- effective removal of analog artefacts (like rainbows and crawls in composite video)

- motion estimation and interpolation that works as well as a human brain works. especially at low frame-rates where there's no falling back on frame blending.

- noise reduction that works as well as a brain works (which is to say the brain can't be fooled by the right signals...)

- stenography, and removal of invisible watermarking (this is more applicable to audio - see the "cinavia" thread in the decrypting forum)

- anything to do with interlace is exceedingly unpopular, but good handling of interlace is necessary.

[slavanap]

Mug Funky,
Thank you for the answer!
And what can you say about creating 3-D film from 2-D or stereopair form single image?
Don't users want to do that on their PCs?

[4evrplan]

Mug Funky,
And what can you say about creating 3-D film from 2-D or stereopair form single image?
Don't users want to do that on their PCs?

They can, if they're willing to manually select areas and assign a depth (or gradient of depths) in every single frame. I can't see a plugin ever doing that. There's just not enough depth information in a video for a computer to do it.

I had an idea for this where I thought maybe the algorithm could look at the relative bluriness of neighboring pixels to get an idea of that pixel's depth (because the optics of a camera progressively blur objects further away and closer to the focal point). However, after reading several technical papers, I've come to the conclusion that it's just not enough. (Truthfully, I didn't understand most of the math, but you don't have to in order to get the point that more is needed to determine depth with any kind of accuracy and precision.)

[slavanap]

Mug Funky,
- noise reduction that works as well as a brain works (which is to say the brain can't be fooled by the right signals...)

There are some of my suggestions:
http://compression.ru/video/denoising/index_en.html
http://www.risingresearch.com/en/denoise/
Describe in more details, please, what filter do you want to see.

- motion estimation and interpolation that works as well as a human brain works. especially at low frame-rates where there's no falling back on frame blending.
http://compression.ru/video/motion_estimation/index_en.html
What you want to improove here?

- stenography, and removal of invisible watermarking (this is more applicable to audio - see the "cinavia" thread in the decrypting forum)
Sorry, did you meant Steganography? Here is example of smart Steganography in video. http://compression.ru/video/stego_video/index_en.html

And about removing logos: http://compression.ru/video/logo_removal/index_en.html
There are several techniques for removing transparent logos more efficiently, but I'm not sure, have them been realized or not. It fits.

[slavanap]

4evrplan,
There's just not enough depth information in a video for a computer to do it.
Yes, of course it is. But several techniques of getting Depth Map (for example, depth from defocus (http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/FAVARO1/dfdtutorial.html) and depth from motion estimation (http://compression.ru/video/motion_estimation/index_en.html) (if objects on the scene are static)) have been already released and, I hope, soon become available for free. But how many people do need this filter?
This subject also fits, thank you. (I'm looking for a subject for free filter I'm going to develop)