OpenCV: Invert a mask?
Question:
Is there a simple way to invert a mask using OpenCV? For example, if I’ve got a mask like this:
010
111
010
I’d like to invert it and get this:
101
000
101
Note: I’m using OpenCV’s Python bindings, so while it would be possible to simply loop over each element in the mask, execution speed could become an issue.
Answers:
If you have an 8-bit mask, then you should do mask = 255 - mask. cv::Mat subtraction operator is overloaded to do scalar per-element subtraction.
For an 8 bit mask using 255 as “on” value:
mask = cv::Mat::ones(mask.size(), mask.type()) * 255 - mask;
I’m using this one instead of Matt M solution as I’m still using OpenCV 2.1.0 for one of my projects.
cv2.bitwise_not(mask) would help here
I think something like this might handle the case where the input mask may have various non-zero values:
cv::Mat1b inputMask = ....;
cv::Mat1b invertedMask(inputMask.rows, inputMask.cols);
std::transform(
inputMask.begin(), inputMask.end(), invertedMask.begin(),
std::logical_not<uint8_t>()
);
Is there a simple way to invert a mask using OpenCV? For example, if I’ve got a mask like this:
010
111
010
I’d like to invert it and get this:
101
000
101
Note: I’m using OpenCV’s Python bindings, so while it would be possible to simply loop over each element in the mask, execution speed could become an issue.
If you have an 8-bit mask, then you should do mask = 255 - mask. cv::Mat subtraction operator is overloaded to do scalar per-element subtraction.
For an 8 bit mask using 255 as “on” value:
mask = cv::Mat::ones(mask.size(), mask.type()) * 255 - mask;
I’m using this one instead of Matt M solution as I’m still using OpenCV 2.1.0 for one of my projects.
cv2.bitwise_not(mask) would help here
I think something like this might handle the case where the input mask may have various non-zero values:
cv::Mat1b inputMask = ....;
cv::Mat1b invertedMask(inputMask.rows, inputMask.cols);
std::transform(
inputMask.begin(), inputMask.end(), invertedMask.begin(),
std::logical_not<uint8_t>()
);