Generating a HSV Histogram without For-loops

Hi,

I'm creating a video processing tool in MatLab, which works, but my code is extremely slow, which I believe is due to my use of for-loops.

One example of this is in creating a HSV Histogram for a video. The code I have is as follows:

function [hsvHist] = generateHSVHist(hsvVid, numRows, numColumns, numFrames)

    H = zeros(12,1);
    S = zeros(4,1);
    V = zeros(4,1);

    for frame = 1 : numFrames        
        for row = 1 : numRows
            for column = 1 : numColumns
                h = hsvVid(row,column,1,frame);
                s = hsvVid(row,column,2,frame);
                v = hsvVid(row,column,3,frame);

                idxH = ceil(h*12);
                idxS = ceil(s*4);
                idxV = ceil(v*4);

                %ensure no zero indices
                if idxH == 0
                    idxH = 1;
                end
                if idxS == 0
                    idxS = 1;
                end
                if idxV == 0
                    idxV = 1;
                end

                H(idxH) = H(idxH)+1;
                S(idxS) = S(idxS)+1;
                V(idxV) = V(idxV)+1;
            end
        end
    end

    %Normalise the histograms
    totalPixels = numFrames * numRows * numColumns;
    H = H/totalPixels;
    S = S/totalPixels;
    V = V/totalPixels;    

    %Concatenate the histograms
    hsvHist = [H;S;V];
end

Do you guys have any tips or advice on how I can approach this problem without using for-loops? Any help is greatly appreciated.

Thanks,

Gareth

Answer 1

[Walter Roberson]

No loop needed.

h = hsvVid(:,:,1,frame);
idxH = max(1, ceil(h*12));
H = histc(idxH(:), 1:12);

[Gareth]

Thanks, this worked perfectly. The function is now roughly 10x faster.

Here's the final code for anyone wondering:

function [hsvHist] = generateHSVHist(hsvVid, numRows, numColumns, numFrames)

    H = zeros(12,1);
    S = zeros(4,1);
    V = zeros(4,1);

    %Histogram for Hue
    h = hsvVid(:,:,1,:);
    idxH = max(1, ceil(h*12));
    H = H + histc(idxH(:), 1:12);

    %Histogram for Saturation
    s = hsvVid(:,:,2,:);
    idxS = max(1, ceil(s*4));
    S = S + histc(idxS(:), 1:4);

    %Histogram for Value
    v = hsvVid(:,:,3,:);
    idxV = max(1, ceil(v*4));
    V = V + histc(idxV(:), 1:4);

    %Normalise the histograms
    totalPixels = numFrames * numRows * numColumns;
    H = H/totalPixels;
    S = S/totalPixels;
    V = V/totalPixels;    

    %Concatenate the histograms
    hsvHist = [H;S;V];
end