UPDATE for clarification-- This is more correctly called Perspective or Keystone Correction as noted by Rich Shumaker.

A deskew filter would allow us to "straighten out" a camera feed that is bringing in a scoreboard video feed which would make it easier to integrate it with our sports streams.

MimoLive Deskew Filter

Maybe this will clarify what I am asking

MimoLive Deskew Filter.png (1,202kb) downloaded 0 time(s).

Note that we do not intend to use this to bring in the actual scoreboard image into the production but would only use the corrected video to capture portions of the scoreboard with the changing data (mostly the clock) to make for a better viewer experience. Also the UVMap is an OK idea but we would need to generate a new UVMap at every game since the camera will never be in the exact same place twice.

Also I have found an example with Python that someone who was handing might be able to further develop.

Python Sample Code

To make this work, I think adding horizontal and vertical skew to the position settings plus the ability to crop post position changes would get it done.

I like the idea .... but imo most users like scoreboard data to be pulled from the scoreboard display system to display them in custom titles
Just looks so much nicer than a PIP of the board.

But until such filter is available and dont want to use scoreboard OCR an idea might be
:

In order to display such scoreboard (video)image the camera is fixed during the whole production. As such one can:
Use a basic Virtual set with a UVMap
load a fulle screen UVmap in Gimp
Get a screenshot from the scoreboard input and put on top of the UVMap in Gimp
Link both layers and use the perspective tool to adjust the image
Select the part of the scoreboard image and remove/erase the outerside of it (both layers still linked together), or crop later using a Mix Input
Export the UVmap layer of it to your virtualset folder (same name as "normal" original uvmap)
Load the virtual set and set the UVmap layer to the scoreboardfeed, you now have a deskewed view of the video (maight want to set the MAIN layer of the virtual set to full transparent before overlaying


Screenshot 2023-11-02 112641.jpg (196kb) downloaded 1 time(s).