I'm not sure if I'm doing something wrong, but I have a strange kind of "misunderstanding" with opencv function (in python) for random number generation - randn() http://docs.opencv.org/2.4/modules/core/doc/operations_on_arrays.html#randn
When I try to add gaussian noise to RGB image (adding normally distributed random numbers in "dst" matrix that has 3 channels), those random numbers get only distributed through one channel (the first one... blue). Other channels stay unchanged. I am implementing simple peace of code
cv2.randn(gauss, mean, sigma)
which with
bg, gg, rg = cv2.split(gauss)
shows that I haven't changed anything in gg and rg.
Am I doing something wrong?