Transforming a square image to an image of curved sides

Hi all!

I'd be very thankful for any tip on this!

Challenge:

I have a square picture of the moon (1024x1024 pixels). The 4 corners all have different coordinates (latitude and longitude). I have interpolated the coordinates for the image every 8 pixels and I can creat a matrix with that info.

My solution:

I though of interpolating all the coordinates for all the pixels and then put the pixel in the coordinate matrix but that takes me an hour to compute an image... and I have several hundred thousands to process.

Question:

Is there any other method to transform a square image into another image with non linear sides?

Thank you all!

The code here:

     for x=1:129
        for y=1:121
                inputPointsX(y,x)=(coords{(x-1)*121+y,4}-mapCoord(4,2))*xScale;
                inputPointsY(y,x)=(coords{(x-1)*121+y,3}-mapCoord(4,1))*yScale;
        end
     end

    xcoords = 1:129;
    ycoords = 1:121; 
    [X,Y] = meshgrid(xcoords,ycoords);   

    try 

    for x=1:1024      % Interpolated version - SLOW!!!
         for y=1:1024-64
             xccc = interp2(X,Y,inputPointsX,x/8+1,y/8+1);
             yccc = interp2(X,Y,inputPointsY,x/8+1,y/8+1);
             Image(round(yccc),round(xccc))=I(y,x);
         end
    end

Answer 1

[Sean de Wolski]

Part 5 of:

showdemo mapexgeo

Maybe? Does the code you have right now work as you expect but is just slow? If it is, post it, and maybe we can chop away at making it faster.

[Luis]

Thank you Sean! I added the code in the question.... it's those fors that are slowing down the whole process.

[Luis]

the 1024x1024 is not accurate since I had to remove some lines of pixels.

[Luis]

oh and mapexgeo, I've been there but with the kind of data I have I don't see much use for it.