TDA4VM: About the bit transform LUT table of the TDA4 SDK0805 VISS-FCP module

I noticed that when the data enters the FCP module, there will be a 16bit→12bit LUT table to convert the input raw16 into raw12, and then pass the raw12 to CFA. But looking at the SDK code, the LUT table of 16to12 calls the LUT table of 20to16, and the state in the default assignment function is disabled, so I have a question, what is the bit conversion relationship in this place, and how many bits of the image are entered into the CFA module？