Fastest way to get number of white pixels in a binary image using OpenCV

Question

What is the fastest way to get number of white pixels in a binary picture using OpenCV? Is there something faster than using two for loops and accessing the image pixel by pixel?

Answer 1

The most succinct way to accomplish this is:

cv::Mat image, mask;    //image is CV_8UC1
cv::inRange(image, 255, 255, mask);
int count = cv::countNonZero(mask);

If you are operating on a binary image, then the call to cv::inRange() is unnecessary, and simply cv::countNonZero() will suffice.

Although any method must iterate through all the pixels, this may be able to harness OpenCV's built-in parallel_for_(), which allows parallel execution.

If your image is continuous, you can iterate through all the data using a single loop.