Time complexity of incrementing a value in a matrix in my code is awful.

I'm writing a function that takes in two images, looks at respective pixel pairs between the images (e.g. [image_A(1,1),image_B(1,1)], [image_A(1,2),image_B(1,2)]), defines a binning scheme for these joint values, and returns a matrix that indicates how many pixel pairs fell into the bins. I'm doing this using histc to determine which bin each pixel falls into independent of the other, then updating the bin matrix considering the two values jointly as I don't know if I can do this with histc on two values. The code is taking forever to execute and it appears to be this one simple line indicated by the MATLAB profiler. See code below:

    function [joint_dist] = jointDistFromImages(image_A, image_B, num_bins)
    %%Map joint pixel values for image_A and image_B into an n*n bin matrix 

        num_pixels = numel(image_A);
        image_linear_A = reshape(image_A,1,num_pixels);
        image_linear_B = reshape(image_B,1,num_pixels);
        joint_dist = zeros(num_bins, num_bins);
        increment = (1/num_pixels);

        for i=1:num_pixels
            temp1 = histc(image_linear_A(i),0:(256/num_bins):256);
            temp1 = temp1(1:1:end-1);
            temp2 = histc(image_linear_B(i),0:(256/num_bins):256);
            temp2 = temp2(1:1:end-1);
            row = find(temp1==1);
            col = find(temp2==1);
            joint_dist(row,col) = joint_dist(row,col) + increment;   
        end  

    end

Here is the output of the MATLAB profiler:

%

Answer 1

[Walter Roberson]

http://www.mathworks.com/matlabcentral/fileexchange/9896-2d-histogram-calculation