MATLAB: Normalize hist3 plot

hist3normalize

Is it possible to normalize a hist3 plot so the vertical axis does not have counts but rather probabilities so the volume under the surface integrates to 1? For example,

X = sortrows(randi(10,100,2));
hist3(X)

results in a bivariate histogram of the columns of X but the vertical axis is counting the number of times each combination occurs. Instead, is it possible to change the vertical axis so it represents the fraction of times a combination occurs?

Thanks.

Best Answer

Use histogram2 instead of hist3. As stated in the "Alternative Functionality" section of the hist3 documentation page, histogram2 allows you to specify the 'Normalization' parameter.

Related Solutions

MATLAB: For texture classification i have taken lbp as a feature..after applying lbp i have taken histogram..how to normalise the bins of histogram..and what is meant by normalise bins of histogram

Normalization would mean that you either divide by the max value of the counts, or that you divide by the sum of all the counts.

[pixelCounts grayLevels] = imhist(lbpImage, 256);
normalizedHist1 = pixelCounts / max(pixelCounts(:));
normalizedHist2 = pixelCounts / sum(pixelCounts(:));

It means that either the max value of any count will be 1.0, or that the sum of all counts would now be 1.0, depending on which method you use.

MATLAB: Normalized distribution for histogram

If you have the Statistics Toolbox, you might find the dfittool distribution fitting tool quite useful.

doc dfittool

If not, you can normalize a histogram by scaling the counts in each bin. With the normalized counts, you can plot both the normalized histogram and your curve. The trick is to identify the appropriate scaling factor.

Here is some example code where I plot the normal probability with the normalized histogram data:

% Arbitrary data generation and statistics
dataVec = randn(1000,1);
muData = mean(dataVec);
stdData = std(dataVec);
% Defining the bin centers
binStep    = 0.1;
binCenters = -5:binStep:5;
% Compute the histogram scaling factor. (If all histogram counts are in a
% single bin, the probability = 1)
binCount   = histc(dataVec,binCenters);
probScale  = sum(binCount)*binStep;
% Plot the normalized histogram and change the bar color for line
% visibility
histHandle = bar(binCenters,binCount/probScale,'hist');
set(histHandle,'FaceColor',[1,1,1]); 
hold on;
% Overlay the distribution fit on the histogram
x = binCenters;
y = normpdf(x,muData,stdData);
plot(x,y);
xlabel('Data');
ylabel('Probability');

Hope this helps to get you started!

Best Answer

Related Solutions

MATLAB: For texture classification i have taken lbp as a feature..after applying lbp i have taken histogram..how to normalise the bins of histogram..and what is meant by normalise bins of histogram

MATLAB: Normalized distribution for histogram

Related Question