I don't know what software package you are using. Result output style can vary software to software. In short, if your optimized structure is either a local minima or global minima, all of the normal modes of vibrations will be real and not imaginary. If your structure is not a minima (either global or local) rather a saddle point (like Transition state) one or more of the normal modes will be imaginary (for transition state only one normal modes of vibrations will be imaginary). This way you can separate between minima and saddle points.
However, to best of my knowledge it is not possible to separate local and global minima by observing the normal modes (real or imaginary) because for both of them all the normal modes of vibrations will be real. If you want to find out local and global minima and distinguish them, you need to scan through all minima by means of some automated algorithm like basin hopping or particle swarm based optimization in combination with QM software like VASP or Gaussian.
PS. Some software like Gaussian do report imaginary freq as negative number e.g -450.31 and not as 450.31i
Exactly as nicely stated above by Turbasu, an imaginary frequency is indication of a saddle point. At minima whether they are local or global, all frequencies should be real. You would have to to scan the whole potential energy surface...
If you are using vasp, the you have to add some additional command lines in the INCAR file such as NFREE=2 or 3, POTIM=0.015, IBRION =5. For this I will suggest you to go through https://cms.mpi.univie.ac.at/wiki/index.php/Vibrational_frequencies_of_CO_on_Ni_111_surface#INCAR.
After running this INCAR for your system, you should get an OUTCAR file. In the linux terminal, all you have to type is " grep cm-1 OUTCAR".