CODEWITHSUNDEEP
pythonopencvnumpy
Abid Rahman K
Abid Rahman K

Reputation: 52646

Difference in adding two images in numpy and opencv

when i add two images in opencv, sum is limited to 255. (Both images are of uint8)

ie 175+100 is 255 in opencv.

but if we add it using numpy.add function, result is not a limited one.

ie 175+100 is 19 in numpy.

Question:

1) Why it happen so?

2) Is there a way for np.add to behave like cv2.add? ie limit sum to 255?

Thanks in advance.

Upvotes: 1

Views: 2712

Answers (1)

Travis Oliphant
Travis Oliphant

Reputation: 46

NumPy uses "modulo" arithmetic on overflow rather than clipping. This is the behavior of add on uint8 integers in C. So, 175+100 % 256 = 19 which is the result you are getting.

To get this clipping behavior, you will need to do some work:

Here are a couple of ideas:

1) Use a higher precision: 

im1 = im.astype('u2')
im2 = im.astype('u2')
tmp = im1 + im2
result = tmp.clip(0,255).astype('u1')

2) Make sure the result is >= both the input values:

tmp = im1 + im2
mask = (tmp < im1) | (tmp < im2)
tmp[mask] = 255

Upvotes: 3

Related Questions