%by Dr. Rex Cheung
%cheung.r100@gmail.com
%This programs calculate and visualize the dice similarity (volume overlap) of 2D binary images.
%This program is useful for quantifying the accuracy of 2D image
%registration using DICE similarity (see wikipedia):               
%      Dice Coef = 2*intersect(A,B)/(absolute(A)+absolute(B))
%This program also allows users to see the parts that are not registrated
%to guide further refinement registration. 

function [OverlapImage DiceCoef] = DiceSimilarity2DImage(img1, img2)
%The steps are:
%1. set one image non-zero values as 200
%img1(img1>0)=200;
img1=200*img1;

%2. set second image non-zero values as 300
%img2(img2>0)=300;
img2=300*img2;

%3. set overlap area 100
OverlapImage = img2-img1;

%4. count the overlap100 pixels
[r,c,v] = find(OverlapImage==100);
countOverlap100=size(r);

%5. count the image200 pixels
[r1,c1,v1] = find(img1==200);
img1_200=size(r1);

%6. count the image300 pixels
[r2,c2,v2] = find(img2==300);
img2_300=size(r2);

%7. calculate Dice Coef
DiceCoef = 2*countOverlap100/(img1_200+img2_300);

%8. visualize the volumes by slicing 
%figure(1);image(OverlapImage);colormap(gray);title('Overlapping area used to calculate Dice Coef')

%command
% [a b]=DiceSimilarity2DImage(cropprostmask1,cropprostmask2)