C++ OpenCV equivalent to Matlab imfill? - c++

Question

I have an original image
if I use imfill in Matlab, the result is
But I use the following code in C++ as an attempt to mimic the imfill from Matlab:
cv::Mat BW = cv::imread("smallHoles.bmp", cv::IMREAD_GRAYSCALE);
cv::Mat edgesNeg = BW.clone();
cv::floodFill(edgesNeg, cv::Point(0, 0), cv::Scalar(255));
bitwise_not(edgesNeg, edgesNeg);
cv::Mat filledEdgesOut = (edgesNeg | BW);
cv::imwrite("C:/Users/me/circleCVfill.bmp", filledEdgesOut);
the result is that all the black objects, even the black vertical object, gets filled with white pixels, which doesn't match the Matlab result:
What is the correct C++ OpenCV code to use?
EDIT:
The Matlab code I use:
BW=imread('smallHoles.bmp');
BWfill=imfill(BW,'holes');
EDIT2:
I also tried the following code
cv::Mat circle = cv::imread("smallHoles.bmp", cv::IMREAD_GRAYSCALE);
std::vector<std::vector<cv::Point> > contours;
cv::Mat contourOutput = circle.clone();
cv::findContours(contourOutput, contours, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_SIMPLE);
cv::Mat contourImage(circle.size(), CV_8UC1, cv::Scalar(255));
cv::drawContours(contourImage, contours, -1, cv::Scalar(255));
cv::imwrite("C:/Users/me/output.bmp", contourImage);
but it also gave just the same result of an all-white image just like the previous C++ code I tried
EDIT3:
I tried
std::vector<std::vector<cv::Point> > contours;
cv::Mat contourOutput = circle.clone();
cv::findContours(contourOutput, contours, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_SIMPLE);
cv::drawContours(contourOutput, contours, -1, cv::Scalar(255));
cv::imwrite("C:/me/output.bmp", contourOutput);
It returns as output the original image with the 4 small black holes that should be filled

Answer 1

In trying your EDIT3:
cv::Mat circle = cv::imread(argv[1], cv::IMREAD_GRAYSCALE);
std::vector<std::vector<cv::Point> > contours;
cv::Mat contourOutput = circle.clone();
cv::findContours(contourOutput, contours, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_SIMPLE);
cv::drawContours(contourOutput, contours, -1, cv::Scalar(255));
cv::imwrite("output.bmp", contourOutput);
I got the following:
https://i.stack.imgur.com/T5YBR.png
This is expected. If you check the documentation on the function drawContours, it says
thickness – Thickness of lines the contours are drawn with. If it is negative (for example, thickness=CV_FILLED ), the contour interiors are drawn.
So we can see the desired shapes are taken but the default for drawing is a thickness of one, but you want a negative value (or CV_FILLED) as the next parameter and you should get your desired solution.
I haven't used the floodFill function though, so can't help you there though.