python - Evaluate gradient angle at every pixel of a fringe pattern image -

i trying evaluate gradient angle @ every pixel location in fringe pattern.the problem there pixel locations gradient goes 0 , @ these points angle calculated erroneous.i have tried interpolation changes angle values @ pixels correct.i need accurate gradient angle values in computation of other quantities.is there way obtain accurate gradient angles without having face above problem?if not possible,is there way interpolation in erroneous region?or other way overcome problem? implementation using opencv , python:

k1  = np.array([[-1,0,1],[-1,0,1],[-1,0,1]]) dx1 = cv2.filter2d(image,cv2.cv_64f,cv2.flip(k1,-1),(-1,-1)) dy1 = cv2.filter2d(image,cv2.cv_64f,cv2.flip(np.transpose(k1),-1),(-1,-1)) alpha1 =np.arctan2(np.abs(dy1),np.abs(dx1)) 

i testing on simulated circular fringe pattern added gaussian random noise.i have attached link image without added noise. https://drive.google.com/file/d/0b7tjuyo9bt28zuq4ufnxqwnhzwc/view?usp=sharing

edit 1: trying find direction of fringes in image direction of tangent @ every pixel.so finding gradient first , using fact tangent perpendicular gradient.so if can suggest alternative method find tangent,it welcome.