Fiber orientation - atan2 problem

Hello,

I am writing a program allowing me to plot angular orientations of fibers from microscope images.

I am processing the images with the blocproc function. The angles are calculated from the structure tensor:

ang=.5*atan2(2*Ixx*Iyy,Iyy*Iyy-Ixx*Ixx)*180/pi;

where Ixx and Iyy are the structure tensor components.

In the resulting plot however, I notice a peak at 45°, i.e. for arctan(..)=90°. Every other angle seems fine, I am wondering whether I am dealing with a problem for the atan2 function. I understand that tan is undefined for 90°. I know for certain that there must be a problem somewhere, as when I rotate the source image by 90°, I again get a peak at 45°.

Could it be a numerical calculation problem?