Open nlolnlolnlolnlo opened 4 months ago
The ntsc-rs comb filter essentially just averages the luma from successive fields to demodulate them. I'm not sure how actual comb filters work but at least one source suggests this is a known problem with two-line comb filters. I'm not sure why this wouldn't also apply to rudimentary one-line comb filters.
Fixing this might require implementing some sort of adaptive comb filter, which I might eventually dig into (no promises).
it might be worth looking into some of the work that's been done for cvbs-decode/vhs-decode. i haven't really looked into their chroma decoder options but i believe they have some sort of implementations of 2d/3d comb filtering.
https://github.com/oyvindln/vhs-decode & https://github.com/oyvindln/vhs-decode/wiki/CVBS-Composite-Decode
the new interlacing features fix a lot of my past suggestions to improve ntsc-rs, but something i've noticed with the new option is that the comb filter leaves the output fields much fuzzier than a real comb filter. is this something that can be fixed?