Daniel Armyr

Hi.
That is a pretty good idea. I will take it into consideration, but for now, here is a quick hack that will give you what you are looking for.

Use it the same way that you use the original function, only add a final argument (scaleFactor) that tells the function what scale factor to apply. If you want a density function, you set the last argument to 1/numel(X). In order to avoid extensive recoding, you will have to provide all the other arguments in order to use a scaling factor.

Example of density plot of random data
nPoints = 1e5;
clpoudPlot ( randn(nPoints,1), randn(nPoints,1), [], false, [], 1/nPoints );

--

function [ varargout ] = cloudPlot( X, Y, axisLimits, useLogScale, bins, scaleFactor )

% Check the data size
assert ( numel(X) == numel(Y), ...
    'The number of elements in X and Y must be the same.' );

if ( nargin >= 3 && ~isempty(axisLimits) )
    pointSelect = X<=axisLimits(2) & X>=axisLimits(1) & ...
        Y<=axisLimits(4) & Y>=axisLimits(3);
    X = X(pointSelect);
    Y = Y(pointSelect);
    axisLimitsSet = true;
else
    axisLimitsSet = false;
end

if ( nargin < 4 || isempty(useLogScale) )
    useLogScale = false;
end

if ( nargin < 5 )
    bins = [];
end

  
%Remove any nans or Infs in the data as they have no meaning in this
%context.
pointSelect = ~(isinf(X) | isnan(X) | isinf(Y) | isnan(Y));
X = X(pointSelect);
Y = Y(pointSelect);
    
% Plot to get appropriate limits
h = [];
if ( axisLimitsSet )
    g = gca;
    set ( g, 'Xlim', [axisLimits(1) axisLimits(2)] );
    set ( g, 'Ylim', [axisLimits(3) axisLimits(4)] );
    set ( g, 'units', 'pixels' );
else
    h = plot ( X(:), Y(:), '.' );
    g = get( h, 'Parent' );
end
xLim = get(g, 'Xlim' );
yLim = get(g, 'Ylim' );

%Get the bin size.
unitType = get(g,'Unit');
set(g,'Unit','Pixels')
axesPos = get(g,'Position');
nHorizontalBins = axesPos(3);
nVerticalBins = axesPos(4);
set(g,'Unit', unitType );

% Clear the data, as we actually don't want to see it.
if ( ~isempty(h) )
    set ( h, 'XData', [] );
    set ( h, 'YData', [] );
end

% Allocate an area to draw on
if ( isempty(bins) )
    bins = ceil([nHorizontalBins nVerticalBins ]);
else
    assert ( isnumeric(bins) && isreal(bins) );
    assert ( numel(bins) == 2 );
end
binSize(2) = diff(yLim)./(bins(2));
binSize(1) = diff(xLim)./(bins(1));

canvas = zeros(bins);

% Draw in the canvas
xBinIndex = floor((X - xLim(1))/binSize(1))+1;
yBinIndex = floor((Y - yLim(1))/binSize(2))+1;

% Added security: Make sure indexes are never outside canvas. May not be
% possible.
pointsSelect = xBinIndex > 0 & xBinIndex <= bins(1) & ...
    yBinIndex > 0 & yBinIndex <= bins(2);
xBinIndex = xBinIndex(pointsSelect);
yBinIndex = yBinIndex(pointsSelect);

for i = 1:numel(xBinIndex);
    canvas(xBinIndex(i),yBinIndex(i)) = ...
        canvas(xBinIndex(i),yBinIndex(i)) + 1;
end

% Show the canvas and adjust the grids.
if ( useLogScale )
    h = imagesc(xLim, yLim,log10(canvas*scaleFactor)');
else
    h = imagesc(xLim, yLim, canvas'*scaleFactor);
end

axis ( 'xy' );
axis ( 'tight' );
set ( g, 'units', 'normalized' ); % Enable resizing

% Optionally return a handle.
if ( nargout == 1)
    varargout{1} = h;
end

Added the source code and rerendered the film to remove some glitches.

Of course you are right. New version with better scaling, support for hold states, and support for manual setting of transparency. See the published demo file for examples.

Great Code Daniel.

Can you please suggest how I can obtain probability density instead of histogram in your Code or How I can divide the frequency in all the bins with a constant number.

Thank you

Sireesh Dadi

nice