# Average transformation matrix?

I have several estimates of the transformation matrix between two planes and some values that give some indication of the error involved in the estimate.

How can I use this information to gain the best average case estimate of the transformation matrix?

Could I simply average all the matrices? If so, how would I go about it?

Or would it be best to assign a probability proportional to the accuracy and include that in a model to estimate it or something along those lines....

...any ideas?

-