getHSVHistogram.m

function [Hist] = getHSVHistogram(imageName)

% read RGB data:

RGB = imread(imageName);
RGB = rgb2hsv(RGB);

% get image size:
[M,N,~] = size(RGB);

range = 0.0:0.1:1.0;

Hist = zeros(length(range),length(range),length(range));

for i=1:M
    for j=1:N        
        nn1 = round(RGB(i,j,1) * 10)+1;
        nn2 = round(RGB(i,j,2) * 10)+1;        
        nn3 = round(RGB(i,j,3) * 10)+1;
        
        Hist(nn1, nn2, nn3) = Hist(nn1, nn2, nn3) + 1;
        
    end
end

Hist = Hist / (M*N);

HSVcomp.m

function [Similarity] = HSVcomp(Hist1, Hist2)

range = 0.0:0.1:1.0;
rangeNew = 0.0:0.05:1.0;
[x,y,z]    = meshgrid(range);
[x2,y2,z2] = meshgrid(rangeNew);

% decision thresholds:
t = 0.010;
t2 = 0.8;

Hist1 = interp3(x,y,z,Hist1,x2,y2,z2);
Hist2 = interp3(x,y,z,Hist2,x2,y2,z2);
DIFF = abs(Hist1-Hist2) ./ Hist2;

DIFF(isnan(DIFF)) = 0; 
DIFF(isinf(DIFF)) = 1; 

DIFF2 = DIFF(Hist1 > t & Hist1 < t2);
% keep distance values for which the corresponding query image's values
% are larger than the predefined threshold:    
DIFF = DIFF(Hist1>t);

% keep error values which are smaller than 1:
%DIFF2 = DIFF(DIFF<t2);
L2 = length(DIFF2);

Similarity = length(DIFF) * mean(DIFF2)/ (L2^2);

why they have multiply by 10 and add by 1 in round off function.