MATLAB: getting rid of a for loop - matlab

Question

I'm looping over a vector and calling a function for every element in it
I want to get rid of the loop, I can't see how.
a short description about the code:
I have imageA and ImageB, I want to map ImageB into ImageA where the user chooses 4 points in ImageA that locates the position of the pattern (ImageB)
this is the part of code that has the for loop
newIm=im1; % newIm will have the mapped image
[r,c]=size(im1); %size of imageA
[p1,p2]=meshgrid(1:c, 1:r); % using indexes matrixes
% p1 has the X indexes
% p2 has the Y indexes
z = ones(r,c); % a vector of ones (for mapping algorithm)
% now calculate for every point in imageA the indexes for its new value
% from imageB ( Inverse Mapping)
% X a b e X'
% Y = pinv c d f * Y'
% 1 g h 1 1
newmatrix=pinv(mat)*[p1(:) p2(:) z(:)]';
newX=newmatrix(1,:)./newmatrix(3,:); % devide the x with z
newY=newmatrix(2,:)./newmatrix(3,:); % devide the y with z
newX=round(newX);
newY=round(newY);
% %% the indexes we want are the ones in the range of ImageB
Xindex=(newX<Br & newX>1); % this gives us a vector of zeros and ones
Yindex=(newY<Bc & newY>1);
ourPixels=Xindex(:).*Yindex(:); % gives us a vectors that has ones
% in the places where the Y and X are both in the range
ourIndexes=find(ourPixels==1); % a vector that holds the indexes that we want
IndexesSize=size(ourIndexes);
XXX=newX(ourIndexes); %the indexes we want
YYY=newY(ourIndexes);
% loop the quadrilateral and "paint" it (only looping the quadrilateral
% and not the whole image)
for i=1 : IndexesSize(1)
newIm(ourIndexes(i))=BilinearInterpolation(im2,YYY(i), XXX(i));
end
fprintf('press any key to see the result...');
pause;
imshow(newIm);
title('The new mapped Image');
this is the function
function NewVal=BilinearInterpolation(imgB,x,y)
% inpust: ImageB- the pattern image
% x and y the coordinates that locates the inverse mapping
% returns the new color of a spicifc pixel (x',y') in imageA
% SW SE NE NW are neighboors of (x,y) which will give us the new color
SW=imgB(floor(x),floor(y)+1);
SE=imgB(floor(x)+1,floor(y)+1);
NE=imgB(floor(x)+1,floor(y));
NW=imgB(floor(x),floor(y));
% the equations are the same as in the lecture
% S = SE * delta(x) + SW*(1- delta(x))
% N = NE * delta(x) + NW*(1- delta(x))
% V = N * delta(y) + S*(1 - delta(y))
deltax=x-floor(x);
S=SE*deltax+SW*(1-deltax);
N=NE*deltax+NW*(1-deltax);
deltay=1-(y-floor(y));
NewVal=N*deltay+S*(1-deltay);
end
I tried this:
newIm(ourIndexes)=BilinearInterpolation(im2,YYY, XXX);
but of course it didn't work
I tried changing the function to work with (.*) but also had a lot of problems

Answer 1

With a little modification you can make your function accept an array of points.
Firstly as mentioned in the comments change your * to elementwise multiplication .*.
Also you'll need to change the times where you access imgB so that it will accept a list of points. Inputting two vectors (i.e. using the code unchanged) will call all combinations of those points i.e nxn points. Normally I would just use a for loop to avoid this as I am lazy and rarely need to optimise my code that much.
As you specifically don't want to use a loop, the easiest way I know to do this is to replace the calls to the subscripts with the appropriate linear indices using sub2ind. For example, SW=imgB(sub2ind(size(imgB),floor(x),floor(y)+1));
Other than that check that all your vectors have the same long dimension, I think they should but it depends on your input.