How to deinterlace an object within a frame while also using IVTC?

[LaserBones]

The intro to a show I'm encoding is telecined but there is a TV in the background which is interlaced. The interlacing is superimposed on top of the tv so it's field align with the video. I've included a clip. How would I go about properly encoding this. Does decombing target areas of a frame? I don't know where to start.

[Bwaak]

For 24fps content with 30i effects I render two videos: 24p and 60p, then drop both of them in my NLE in two tracks, one below another, set project to 24p, and for the parts with video effects I either use 60p video, or if the effect takes only a part of the frame, I cut out a window in the 24p frame for 60p effect to see through. I render the whole project into 24p.

If the effects are too frequent, I just deinterlace the whole video into 60p.

[lollo]

Did not check the output for fields architecture integrity, and probably something will be wrong when the routines deinterlace the "TV" and then are forced to "repeat the rest of the picture", but something like:

TDeint(mode=0, full=false, cthresh=9, chroma=false, MI=16, edeint=QTGMC(FPSDivisor=2), emask=TMM(mode=0), slow=2)

at first view produce a "good looking" result.

Here the comparison c.avi

[jagabo]

TFM will deinterlace any remaining interlaced portions of the frame after field matching. But you can't get smooth motion of both the 60p and 24p sections unless you encode at 120p (and in most cases you won't see that smooth motion at playback anyway) or variable frame rate.

Here's a side by side comparison of a portion of the clip after IVTC to 24p and 3:2 duplicates to 60p on the left (how the 24p would usually be displayed on a 60 Hz monitor/TV) and a 60p deinterlace on the right.