How to detect shadow using bimodal histogram splitting?

so, here is the algorithm :

but I got confused on step 3. I used another paper but I don't get it too, and this is the explanation

and this is my code so far

I=imread('shadow.tif');
imshow(I);
title('original image','fontsize',14);
T=mean2(I);%mean of the image as threshold
err=T;
while err~=0
  %Divide the image into two using T
  R0=I (I<T);
  R1=I (I>=T);
  %calculate average of each segment
  miu0=mean2(R0);
  miu1=mean2(R1);
  temp=T;
  %Recalculate T
  T=(miu0+miu1)/2;
  err=T-temp;
end

then, how I code step 3 for my program? help me please...