在Matlab中比较图像中RGB值的直方图

Question

So I want the RGB values of an image placed into an histogram and then that histogram will be compared to other image's histogram. Currently this is the code:

if (size(cimg, 3) ~= 3) 
error('rgbhist:numberOfSamples', 'Input image must be RGB.') 
end
nBins = 256;
rHist = imhist(cimg(:,:,1), nBins); 
gHist = imhist(cimg(:,:,2), nBins); 
bHist = imhist(cimg(:,:,3), nBins);
hFig = figure;
%figure 

subplot(1,2,1);imshow(cimg) 
subplot(1,2,2);
hold on
h(1) = stem(1:256, rHist); %hold on 
h(2) = stem(1:256 + 1/3, gHist, 'g'); 
h(3) = stem(1:256 + 2/3, bHist, 'b'); 
hold off
set(h, 'marker', 'none') 
set(h(1), 'color', [1 0 0]) 
set(h(2), 'color', [0 1 0]) 
set(h(3), 'color', [0 0 1]) 
axis square

The code outputs the image along with its RGB histogram value, how can I use that histogram to compare it with other histograms so that I could potentially classify the image as having nearly the same colors as that of another image?

Answer 1

You could use Kullback Leibler Divergence to calculate the distance between 2 histograms.
This is easy as you can treat the Histogram as a distribution.

Since the KL Divergence isn't symmetric one could compute it twice (Namely [X, Y] and [Y, X]) and take the average.