Difference between Cross-correlation and Coherence based registration

I would like to understand the difference between two types of fine registration methods while coregistering

coherence based registration and cross- correlation based registration.
Kindly guide.

I also would like to understand this if anyone has the answer. I am trying to do a multi temporal coherence generation with a stack of TSX images but my stack comes out blurry when I use the coherence based registration. I am in the process of testing with the cross-correlation based registration now.
Thanks in advance.

Your later use of the data does not determine the coregistration technique. You can compute coherence from a stack even it was not created using cccoherence based coregistration. Cross correlation is more common and works most of the times. There are cases when other methods are more applicable, for example you use DEM based coregistration when orbit information is bad or you have extreme shifts between both images.