Apply mask to image with OpenCv Python

Question

i'm trying to use opencv with python and I have this problem:

I have an image and a binary mask (single channel image with 0s and 255) I want to iterate each pixel of the mask and perform some operations on the original image based on the value of the masks pixel. How can I use the numpy optimization to do that?

For example, suppose I want to create a new image where each pixel remains the same if its value in the mask is 0, or its set to (0,0,255) if the pixel in the mask is 255, like:

def inpaint(originalImage, mask):
    [rows, columns, channels] = originalImage.shape
    result = np.zeros((rows,columns,channels))
    for row in range(rows):
        for column in range(columns):
            if(mask[row,column]==0):
                result[row,column] = originalImage[row,column]
            else:
                result[row,column] = (0,0,255)
    return result

How can I optimize this using numpy? Thank you very much

Divakar · Accepted Answer · 2018-04-25 16:59:36Z

7

We can use np.where after extending the mask to 3D that let's it do the choosing in a broadcasted manner -

np.where(mask[...,None]==0, originalImage,[0,0,255])

Or staying closer to the original code, make a copy and then assign in one go with the mask -

result = originalImage.copy()
result[mask!=0] = (0,0,255)

edited Apr 25, 2018 at 16:59

answered Apr 25, 2018 at 16:53

Divakar

222k19 gold badges273 silver badges374 bronze badges

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

5 Comments

Federico Taschin Over a year ago

Thank you very mutch, it seems to work but if I try to use imshow it shows a black image, and if I use resize() I get this error: cv2.error: OpenCV(3.4.1) C:\build\master_winpack-bindings-win32-vc14-static\opencv\modules\imgproc\src\resize.cpp:3922: error: (-215) func != 0 in function cv::hal::resize

Divakar Over a year ago

@FedericoTaschin I think you need to convert to uint8 type with result.astype(np.uint8)

Federico Taschin Over a year ago

Howevr, the second way works! Thank you very much! I have another question: what if I wanted to perform a more complex operation on each pixel, that involves more than a simple assignment, for example calling a function?

Reine_Ran_ Over a year ago

@FedericoTaschin Have you figured out how to perfom the more complex operation on each pixel by calling a function? Mind sharing it here?

Federico Taschin Over a year ago

No, I switched to Java opencv because of some of the operations that were too complex to be done in python without built-in functions

Collectives™ on Stack Overflow

Apply mask to image with OpenCv Python

1 Answer 1

5 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Your Answer

Sign up or log in

Post as a guest

Related