numpy conditionally replace scalar/boolean with array

Question

I have a 3D array (x, y, RGBA) and my goal is :

find which pixels are blank RGBA=[0,0,0,0] then change their color to blue, and for other pixels change color to green.

As far as i see it it can be done in 2 steps :

1- create a 500x500 array with bool True if pixel has value, False if blank

2- then apply a function to replace True by [0,0,255,255] and False by [0,255,0, 255]

after numerous searches (i'm not a python wizard) I managed to achieve 1- in a pythonic way (at least my hope...)

img.shape
>(500, 500, 4)
img_bool = np.equal(img[:,:], [0, 0, 0, 0]).all(axis=2)
img_bool.shape
>(500, 500)

my guess for step 2 was trying such syntax:

img_final = np.where(img_bool, [0,0,255,255], [0,255,0,255])

or

np.choose(img_bool, [[0,0,255,255],[0,255,0,255], out=img_final)

but they give same error (quite logical since both expressions might do the same in fact)

ValueError: shape mismatch: objects cannot be breascast to a single shape

in fact step 2 could be summerized by "how to replace a scalar/boolean by an array/vector in numpy.ndarray ?"

alko · Accepted Answer · 2013-12-20 12:34:44Z

1

For your first task, using the fact that colors are positive integers, you can use

img_bool = img.sum(axis=2)>0

The second you can do with

img[img_bool] = [0, 0, 255, 255]
img[~img_bool] = [0, 255, 0, 255]

Note, that if I get your description right, you'll original expression returns inverse, i.e. you have to change it to

img_bool = ~np.equal(img[:,:], [0, 0, 0, 0]).all(axis=2)

edited Dec 20, 2013 at 12:34

answered Dec 20, 2013 at 12:26

alko

48.7k12 gold badges99 silver badges105 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

comte Over a year ago

very neat syntax. thx, all your answers work perfectly. just one remark : img.sum(axis=2)>0 could return True if pixel array was [0, -15, 15, 0] no ? but in my case it works perfectly since all values are positives. thx again.

alko Over a year ago

@comte yes, it can, I added clarification. I judged from values equal to 255, that your colors are ideed positive.

Collectives™ on Stack Overflow

numpy conditionally replace scalar/boolean with array

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related