Why NaN inf appears when only use simple sum and multiply

Hello, I am trying to use the code to do some image processing. First I randomly sample 2 images to get 2 vectors and try to locate the element position of the vector in the original image. Finally do some simple calculation based on the position and light intensity information. But finally there are always NaN or Inf occurs in the Ix Iy and It vectors. Could someone help me figure out what cause the Nans and Infs? Thanks!

if true
  % Ims1=imread('abc.jpg');
Ims2=imread('def.jpg');
Ims1=double(Im1); % for comparing image
Ims2=double(Im2);
HEIGHT = size(Im1, 1);
WIDTH = size(Im1, 2);
data_num=10000;    %max 27648
ce=zeros(data_num,2);  %store position   ce(1) for row, ce(2) for column
Im1=Ims1';   Im2=Ims2';
Im1=Im1(:);  Im2=Im2(:);
aa=length(Im1);
p = randperm(aa,data_num); 
y1=Im1(p);
y2=Im2(p);
%y1=double(y1);
%y2=double(y2);
for ii=1:data_num
  cc=floor(p(ii)/WIDTH)+1;
  dd=mod(p(ii),WIDTH);
  if dd==0
      dd=WIDTH;
      cc=cc-1;
  end
  ce(ii,1)=cc;
  ce(ii,2)=dd;
end
Ix=zeros(data_num,1);
Iy=zeros(data_num,1);
It=zeros(data_num,1);
G=zeros(data_num,1);
for jj=1:data_num     
  A1=ce(jj,:);     %A1 selected point position
  if jj==1
      B1=ce(2:end,:); %B1 point position except A1
  elseif jj==data_num
      B1=ce(1:data_num-1,:);
  else
      B1=ce([1:jj-1,jj+1:end],:);
  end
%     dist=bsxfun(@hypot,B1(:,1)-A1(1),B1(:,2)-A1(2));
%     out = B1(dist==min(dist),:); % position of the nearest point  out(1) row   out(2) col
%    out=out(1,:);
%     [rr,~]=size(out);
%     for iii=1:rr
%         if out(iii,1)>A1(1)&&out(iii,1)==A1(2) down=out(iii,:); havedown=1; 
%         else havedown=0; end
%         if out(iii,1)>A1(1)&&out(iii,1)>A1(2) rightdown=out(iii,:); haverightdown=1; 
%         else haverightdown=0; end
%         if out(iii,1)==A1(1)&&out(iii,1)>A1(2) right=out(iii,:); haveright=1; 
%         else haveright=0;   end
%     end
% [~,indx]=ismember(out,B1,'rows');
 % B1(indx,:)=[];
  dn=1;rn=1;rdn=1;count=1; K=10;
 while ~(dn>1&&rdn>1&&rn>1)
  %   [n,~] = findNearestNeighbors(B1,A1,K)
  [n]=knnsearch(B1,A1,'k',count*10);% n is the vector of index
  count=count+1;
  for iii=1:length(n)      %all down rightdown right are for position
      if (B1(n(iii),1)>A1(1))&&(B1(n(iii),2)==A1(2)) 
          down(dn,:)=B1(n(iii ),:);         dn=dn+1;   
      end
      if (B1(n(iii),1)>A1(1))&&(B1(n(iii),2)>A1(2))  
          rightdown(rdn,:)=B1(n(iii),:);   rdn=rdn+1; 
      end
      if (B1(n(iii),1)==A1(1))&&(B1(n(iii),2)>A1(2)) 
          right1(rn,:)=B1(n(iii),:);        rn=rn+1;   
      end
  end
  if (count>6)
      break
  end
 end
 if ((dn>1)&&(rdn>1)&&(rn>1))
  [n]=knnsearch(down,A1,'k',1);
  outd1=down(n,:);
  [n]=knnsearch(rightdown,A1,'k',1);
  outrd1=rightdown(n,:);
  [n]=knnsearch(right1,A1,'k',1);
  outr1=right1(n,:);
  [~,dp]=ismember(outd1,ce,'rows');    %find out which row
  [~,rdp]=ismember(outrd1,ce,'rows');
  [~,rp]=ismember(outr1,ce,'rows');  
  Ix(jj)=(y1(rp,1)-y1(jj,1))/(ce(rp,2)-A1(1,2))+(y1(rdp)-y1(dp))/(ce(rdp,2)-ce(dp,2))+(y2(rp)-y2(jj))/(ce(rp,2)-A1(1,2))+(y2(rdp)-y2(dp))/(ce(rdp,2)-ce(dp,2));
  Iy(jj)=(y1(dp,1)-y1(jj,1))/(ce(dp,1)-A1(1,1))+(y1(rdp)-y1(rp))/(ce(rdp,1)-ce(rp,1))+(y2(dp)-y2(jj))/(ce(dp,1)-A1(1,1))+(y2(rdp)-y2(rp))/(ce(rdp,1)-ce(rp,1));
  It(jj)=(y2(jj,1)-y1(jj,1)+y2(dp)-y1(dp)+y2(rdp)-y1(rdp)+y2(rp)-y1(rp))/4;  
 end
  if ((dn>1)&&(rdn<2)&&(rn<2))
  [n]=knnsearch(down,A1,'k',1);
  outd1=down(n,:);
  [~,dp]=ismember(outd1,ce,'rows');
  Iy(jj)=(y1(dp,1)-y1(jj,1))/(ce(dp,1)-A1(1,1))+(y2(dp)-y2(jj))/(ce(dp,1)-A1(1,1));
  Ix(jj)=0;
  It(jj)=(y2(jj,1)-y1(jj,1)+y2(dp)-y1(dp))/2; 
  end
  if ((dn>1)&&(rdn>1)&&(rn<2))
  [n]=knnsearch(down,A1,'k',1);
  outd1=down(n,:);
  [n]=knnsearch(rightdown,A1,'k',1);
  outrd1=rightdown(n,:);
  [~,dp]=ismember(outd1,ce,'rows');    %find out which row
  [~,rdp]=ismember(outrd1,ce,'rows');
  Ix(jj)=(y1(rdp)-y1(dp))/(ce(rdp,2)-ce(dp,2))+(y2(rdp)-y2(dp))/(ce(rdp,2)-ce(dp,2));
  Iy(jj)=(y1(dp,1)-y1(jj,1))/(ce(dp,1)-A1(1,1))+(y1(rdp)-y1(rp))/(ce(rdp,1)-ce(rp,1))+(y2(dp)-y2(jj))/(ce(dp,1)-A1(1,1))+(y2(rdp)-y2(rp))/(ce(rdp,1)-ce(rp,1));
  It(jj)=(y2(jj,1)-y1(jj,1)+y2(dp)-y1(dp)+y2(rdp)-y1(rdp))/3;    
  end
  if ((dn>1)&&(rdn<2)&&(rn>1))
  [n]=knnsearch(down,A1,'k',1);
  outd1=down(n,:);
  [n]=knnsearch(right1,A1,'k',1);
  outr1=right1(n,:);
  [~,dp]=ismember(outd1,ce,'rows');    %find out which row
  [~,rp]=ismember(outr1,ce,'rows');  
  Ix(jj)=(y1(rp,1)-y1(jj,1))/(ce(rp,2)-A1(1,2))+(y2(rp)-y2(jj))/(ce(rp,2)-A1(1,2));
  Iy(jj)=(y1(dp,1)-y1(jj,1))/(ce(dp,1)-A1(1,1))+(y2(dp)-y2(jj))/(ce(dp,1)-A1(1,1));
  It(jj)=(y2(jj,1)-y1(jj,1)+y2(dp)-y1(dp)+y2(rp)-y1(rp))/3; 
  end
  if ((dn<2)&&(rdn>1)&&(rn<2))
  [n]=knnsearch(rightdown,A1,'k',1);
  outrd1=rightdown(n,:);
  [~,rdp]=ismember(outrd1,ce,'rows');
  Ix(jj)=0; Iy(jj)=0;
  It(jj)=(y2(jj,1)-y1(jj,1)+y2(rdp)-y1(rdp))/2; 
  end
  if ((dn<2)&&(rdn>1)&&(rn>1))
  [n]=knnsearch(rightdown,A1,'k',1);
  outrd1=rightdown(n,:);
  [n]=knnsearch(right1,A1,'k',1);
  outr1=right1(n,:);
  [~,rdp]=ismember(outrd1,ce,'rows');
  [~,rp]=ismember(outr1,ce,'rows');  
  Ix(jj)=(y1(rp,1)-y1(jj,1))/(ce(rp,2)-A1(1,2))+(y2(rp)-y2(jj))/(ce(rp,2)-A1(1,2));
  Iy(jj)=(y1(rdp)-y1(rp))/(ce(rdp,1)-ce(rp,1))+(y2(rdp)-y2(rp))/(ce(rdp,1)-ce(rp,1));
  It(jj)=(y2(jj,1)-y1(jj,1)+y2(rdp)-y1(rdp)+y2(rp)-y1(rp))/3; 
  end
  if ((dn<2)&&(rdn<2)&&(rn>1))
  [n]=knnsearch(right1,A1,'k',1);
  outr1=right1(n,:);
  [~,rp]=ismember(outr1,ce,'rows');  
  Ix(jj)=(y1(rp,1)-y1(jj,1))/(ce(rp,2)-A1(1,2))+(y2(rp)-y2(jj))/(ce(rp,2)-A1(1,2));
  Iy(jj)=0;
  It(jj)=(y2(jj,1)-y1(jj,1)+y2(rp)-y1(rp))/2;
  end
  if ((dn<2)&&(rdn<2)&&(rn<2))
  Ix(jj)=0;
  Iy(jj)=0;
  It(jj)=y2(jj,1)-y1(jj,1);
  end
G(jj)=ce(jj,1)*Iy(jj)+ce(jj,2)*Ix(jj);
end
  A=[nansum(Ix.^2),nansum(Ix.*Iy),nansum(G.*Ix);nansum(Ix.*Iy),nansum(Iy.^2),nansum(G.*Iy);nansum(G.*Ix),nansum(G.*Iy),nansum(G.^2)]; % left side of equation 7
  B=[nansum(Ix.*It);nansum(Iy.*It);nansum(G.*It)]; % right side of equation 7
  C=-pinv(A)*B;
  TTC(i)=1/C(3);
end

Answer 1

[Azzi Abdelmalek]

nan is caused by 0/0 and inf is caused by 1/0 for example

[Thomas Bradley]

Thanks for your reply. I will go through the code again.

Answer 2

[Steven Lord]

You can set an error breakpoint that will cause MATLAB to stop when an Inf or NaN is generated. This will allow you to determine which line generated the Inf or NaN.

[Thomas Bradley]

I tried that but it showed me that it occurs in a built-in function

[Walter Roberson]

Which line was the error shown as occurring on?