2D Adaptive Comb (in ntsc-rs)

[rgm89git]

This discussion i will hold for working with 2D adaptive comb in ntsc-rs, as i will search about how i can do this.
This method is used across many analog video decoders, to separate chroma from luma in composite signals.

Here's what i understand (based on searches):

• 2D adaptive look after the selected line and it's 2 neighbours (above and below, if available)
• Pass them, to a check if neighbours have any motion and edge, getting weights for the difference between the line and the neighbours.
• Estimate how similar they are to the selected line.
• Get the chroma value of the line, computing the weighted sum of differences.

Material:
VHS-Decode - comb.cpp
Extron - NTSC Decoding Basics (Part 3)

If anyone has ideas for how this can be implemented in ntsc-rs, tell here.
I will be very thankful, if i helped on this part.
💜 👍

[valadaptive]

I believe the 2D filter doesn't check for motion; that's the 3D filter. Doing motion estimation would involve passing multiple frames in each time, which would be a significant API rework.

Implementing the 2D filter should be easier. Anyone implementing it can't use vhs-decode's source code as a reference, as it's licensed under the GPL.

[rgm89git]

oh, good note right there.

also noticed, for example, that the Comb filter has issues in interlaced content, but using the Both use field mode it can be fixed, however this will not work correctly on interlaced video.
that's why i showed the idea of adaptive comb, as this will work based on fields.

but i think adding a 2D filter will also be a good idea.