MATLAB: C++ Blur function to MATLAB code

cimage processing

Hi i am trying to convert a blur function I implemented in C++ to MATLAB code, I am getting an error when trying to compute the sum of all elements in an array
Index exceeds the number of array elements (1).
Below is the C++ code.
Mat BlurFunc(Mat Grey, int winsize)
{
	Mat Blur = Mat::zeros(Grey.size(), CV_8UC1);
	for (int i = winsize; i < Grey.rows - winsize; i++)
	{
		for (int j = winsize; j < Grey.cols - winsize; j++)
		{
			int sum = 0;
			for (int ii = -winsize; ii <= winsize; ii++)
			{
				for (int jj = -winsize; jj <= winsize; jj++)
				{
					sum += Grey.at<uchar>(i + ii, j + jj);
				}
			}
			Blur.at<uchar>(i, j) = sum / ((winsize * 2 + 1) * (winsize * 2 + 1));
		}
	}
	return Blur;
}
and this is what I am trying to get to work.
grey_img = rgb2gray(img);
y = size(grey_img);
for i=3:(rows-3)
    for j=3:(cols-3)
        A = grey_img(i-2:i,j-2:j);
        sum = sum(A(:));
        y(i,j) = sum/49;
    end
end
imshow(uint8(y));

Best Answer

You have redefined the function name "sum" as name of a variable. After sum=sum(A(:)) the symbol "sum" is a variable. The next time you call "sum(A(:))", A(:) is treated as index.
Solution: Use another name:
for i=3:(rows-3)
    for j=3:(cols-3)
        A = grey_img(i-2:i,j-2:j);
        S = sum(A(:));
        y(i,j) = S / 49;
    end
end
Notes:
  • For the average over 9 elements you have to divide by 9, not 49.
  • The 3 rightmost columns and the 3 rows on the bottom are not considered in your code.
  • Compared with the C++ code, the output is shifted.
  • You have defined y = size(grey_img) , but later on y is the output matrix. rows and columns are not defined.
W    = winsize;             % Shorter name 
[rows, cols] = size(grey_img);
Blur = zeros(rows, cols);   % Pre-allocate
Div  = (2 * W + 1)^2;
for i = W+1:(rows - W)
    for j = W+1:(cols - W)
        Window    = grey_img(i-W:i+W, j-W:j+W);
        Blur(i,j) = sum(Window(:)) / Div;
    end
end
Related Question