Implement Phase Correlation algorithm

Did you convert the floating point numbers from FFT into fixed point numbers?

Be careful by using division and square root with fixed point numbers.

e.g., 0.25 represented using 8 bits and 7 fractional bits is:

0.0100000 -> 32

and 0.57 is

0.1001001 -> 73

0.25/0.57 = 0.438596

if you use a conventional integer divider, you will obtain as result Q= 2 and R=14, so, Q represents 0.015625 not 0.438596.

This calculation can be done if numerator is shifted 7 bits to left:

Q= 01000000000000/01001001 = 00111000 ->56

and Q represents 0.4375, in this case. Further, a rounding circuit could improve precision.

Similar consideration must be taken for square root.