Data types use in Matlab functions

Question

I am new to Matlab. I have this function which I want to calculate the Euclidean distance between two pixels(RGB).

function[distance]=calc_dist(R1, R2, G1, G2, B1, B2)

    if (R1>R2)
        dR=R1-R2;
    else
        dR=R2-R1;
    end

    if (G1>G2)
        dG=G1-G2;
    else
        dG=G2-G1;
    end

    if (B1>B2)
        dB=B1-B2;
    else
        dB=B2-B1;
    end

  sum=uint64(3*dR*dR+4*dG*dG+2*dB*dB);
  disp(sum);
  distance=(sqrt(double(3*dR*dR+4*dG*dG+2*dB*dB));
end

The problem is the displayed value for sum is 255 each time. This must be happening because the variables are of type uint8. How do I change them? I tried to do some sort of casting

sum=uint64(3*dR*dR+4*dG*dG+2*dB*dB);

but I get this error: 'Undefined function 'uit64' for input arguments of type 'uint8'. How should I display the right value for the sum? Thanks. '

Don't even cast, just do the math. I also don't know why each of those terms should be multiplied - you are computing the Euclidean distance, no? — Ben
and for the record matlab has built in dist() which should be able to calculate distances using several metrics — Ben
The function to cast as a 64-bit unsigned integer is uint64; you were just missing the n. Be that as it may, the casting is occurring too late - after you have done the multiplication and not before. You may want to try something like R1=double(R1) etc. for each input variable (tedious) and then set the Euclidean distance as dist = sqrt((R1-R2)^2 + (B1-B2)^2 + (G1-G2)^2). (Not sure where your 2,3 and 4 come in.) — Geoff
@GeoffHayes Tedious, indeed. It gets the job done, though. Thanks. Isn't there any other way? — Tanatos Daniel

Geoff Geoff · Accepted Answer · 2014-06-24T18:59:16

Consider converting your input of 6 variables to one 2x3 matrix, where the first row is the RGB colours from one pixel, and the second row is the RGB colours from the second pixel:

function[distance]=calc_dist(R1, R2, G1, G2, B1, B2)

rgbPixels = [R1 G1 B1; R2 G2 B2];

% cast as double
rgbPixels = double(rgbPixels);

% compute the difference between the rows
rgbDiffs = diff(rgbPixels);

% compute the Euclidean distance
distance = sqrt(sum(rgbDiffs.^2));

This way, you don't have to change your signature and all casting can be done in one line. Try the above and see what happens!

Data types use in Matlab functions

1 Answers