sort whole Mat image and store indices in OpenCV?

Question

This question is not the duplicate of this question

I want to sort the whole Mat image and store the indices similar to the MatLab's [B , Ix] = sort(A);

But in openCV, the sort() and sortIdx() only works for EACH_ROW or EACH_COLUMN .

Problem: So, how can i sort the whole Mat and store the Indices also in openCV?

PS: I want to get the following:

INPUT =

2 0

4 1

dst_index =

1 1

2 2

dst_sorted =

0 1

2 4

Answer 1

There are a solution but it works only if the image is continuous in memory. This is usually the case unless your image is only a ROI of bigger image. You can check that by using function isContinuous . The trick is to create Mat that uses the same memory buffer as your original image but instead of treating it as N rows and M columns, it treat it as 1 row and M*N columns. This can be done by reshape function.

Mat sameMemoryNewShape = image.reshape(1, 1);

Now you can use sort() or sortIdx() on sameMemoryNewShape.

Answer 2

Based on the solution of 'Michael Burdinov' the following should work.

Mat A = (Mat_<uchar>(3, 4) << 0, 5, 2, 5, 2, 4, 9, 12, 3, 12, 11, 1); // example matrix
Mat B; //storing the 1D index result
Mat C = A.reshape(1, 1); //as mentioned by Michael Burdinov
sortIdx(C, B, SORT_EVERY_ROW + SORT_ASCENDING);
for (int i = 0; i < B.cols; i++) //from index 0 to index rows * cols of the original image
{
    int val =  B.at<int>(0, i); //access B, which is in format CV_32SC1
    int y = val / A.cols; //convert 1D index into 2D index
    int x = val % A.cols;
    std::cout << "idx " << val << " at (x/y) " << x << "/" << y
              << " is " << (int) A.at<uchar>(y, x) << endl;
}

Note that the result B is in CV_32SC1 (32bit integer). Not considering that may lead to "strange" behaviour and could result in 1 0 0 0 .