Spaced Ranger said:

Thanks for the Aleks.J's hardware-upgrade info! He went M.I.A. on the 2 forums after some posts and we never found out about the final result of the upgrade. Too bad about The Law Of Diminishing Returns. I hate that law.

BTW, somewhere in that thread (or one of it's references) was mention that the 3D motion adaptive comb filter (as hardware implemented) is practically ineffective. It only works on static images (to make use of temporal filtering, from images before & after) and switches back to 2D filtering when movement is detected. Aside from the theoretical proof offered by test patterns, the picture is almost always moving in the real world and therefore is almost always being 2D filtered. (I hate The Real World law, too.) Hopefully, AntcuFaalb's usage of "pixel motion estimation" will allow the use of the 3D filter all the time. (I like temporal filtering.)

This "movement or no movement" decision isn't made on a per-frame basis. It's not like some frames are 2D and others are 3D*.

These decisions are usually made at the block or pixel level. Think of a ship moving across a static background. You can, in this case, recover the static background via temporal (3D) comb filtering and use other techniques (2D, etc.) for the ship. Most 3D comb filter designs work this way.

* This is certainly possible, but it's generally not a good idea. Frames with tons of movement are mostly 2D comb filtered, but this a heuristic decision rather than a design one. It's also worth mentioning that the old pre-upgrade Bifrost 3D comb filter for Avisynth made per-frame decisions. O_O