Let's say that we have two $3D$ coordinate systems, $A$ and $B$. Now I want to get the coordinate transform matrix $T_{AB}$ such that for any points $p_A$ in $A$, we can get its coordinate in $B$ by $p_B=T_{AB}p_A$.
I have searched and checked ICP algorithm which can help with that. It needs at least $3$ point pairs in $A$ and $B$ to get $T_{AB}$, and actually the more the better.
But now I have different data. I have paired known camera poses $A$ and $B$. Suppose I have a camera $c$. I know its pose $T_{c_A}$ in $A$ and $T_{c_B}$ in $B$, respectively. How to solve $T_{AB}$ with $T_{c_A}$ and $T_{c_B}$?
And what's the least number of camera pairs I need to solve $T_{AB}$?