How to "zero" everything within a masked part of an image in OpenCV

Question

If I have an image (IplImage 8-bit) and a binary mask (which is also an 8-bit IplImage of the same size, where every pixel has a value of either 0 or 255), how can I make every pixel in the image that corresponds with a pixel in the mask with a value of zero have a value of zero, and every pixel in the image that corresponds with a pixel in the mask with any other value (namely 255) have the same value as in the original image?

In other words, anything that is "in the mask area" will keep its original value, and anything outside the mask area will become zero.

Antonio Sesto · Answer 1 · 2013-01-30T22:50:32.753

23

Simplest way, with 'Mat img' (image to be masked, input) and 'Mat masked' (masked image, output):

  img.copyTo(masked, mask)

where 'Mat mask' is a matrix not necessarily binary (copyTo considers elements with zero value). Masked can be of any size and type; it is reallocated if needed.

See the doc.

edited Jan 30 '13 at 22:50

answered Jan 30 '13 at 17:36

Antonio Sesto

2,868
5
33
51

1

This is the best answer. Bitwise and requires you to separate out the channels (say you have a 3 channel image you want masked via a 1 channel image), and each one, and then merge them. This handles all that on its own. – Jdban101 Mar 10 '15 at 07:33
1

In Python: `cv2.copyTo(image_to_mask, mask_array)` – eric Aug 12 '19 at 23:07

score 10 · Accepted Answer · answered Jul 10 '12 at 14:55

10

You can simply use bitwise_and() function.

Check the documentation.

answered Jul 10 '12 at 14:55

Abid Rahman K

51,886
31
146
157

Thank you. This worked. I had to convert the IplImages to Mat in order to use the bitwise_and() function. – Jackson Dean Goodwin Jul 10 '12 at 15:24
@abid rahman K , Masking should always be binary and single channel ? 8 bit ? – AHF Apr 03 '14 at 07:01
To use them in functions like bitwise_and() etc, yes, it should be 8bit single channel binary image. – Abid Rahman K Apr 03 '14 at 07:32

score 3 · Answer 3 · answered Jul 10 '12 at 14:56

3

Multiply or bit-and the mask with the image. There are some OpenCV functions for that, but I do not know their names for the C interface.

in C++:

Mat image, mask;

image = image * mask;
// or 
image = image & mask;

answered Jul 10 '12 at 14:56

Sam

19,708
4
59
82

How to "zero" everything within a masked part of an image in OpenCV

3 Answers3

Linked