determining B frames, telecide() removing wrong dup

[jmac698]

I have a tough problem here, when IVTC a source, I found that on B frames they are partially blended with the following frame, which happened to be a scene change and looked really ugly. If the IVTC could have only picked the previous frame it would look much better. I want to be able to throw out the 2nd to last of each dup frame after deint. Ideally I could also know to do this only with B frames. Any ideas?

[Mug Funky]

ugly solution, but you could try run "improvesceneswitch" before running decimate?

[chipzoller]

You can set the dropped frame cycle threshold using tdecimate with the "cycleR" and "cycle" functions. See the documentation for the correct usage.
I hope this was to what you were referring in your post.