matlab image rotation code - matlab

Question

I'm beginner in computer vision. I'm trying to do a rotate transformation using matlab. My code is
I = imread('Koala.jpg');
rows = size(I, 1);
cols = size(I, 2);
deg = 45;
deg = deg * pi / 180;
C = uint8(zeros(size(I)));
mid = ceil([rows+1 cols+1] / 2);
[x1, x2] = meshgrid(1:rows, 1:cols);
M = [cos(deg) sin(deg); -sin(deg) cos(deg)];
X = bsxfun(#minus, [x1(:) x2(:)], mid) * M;
X = round(bsxfun(#plus, X, mid));
x1 = X(:, 1);
x2 = X(:, 2);
x1(x1<1) = 1;
x2(x2<1) = 1;
x1(x1>rows) = rows;
x2(x2>cols) = cols;
X = [x1(:) x2(:)];
m = 1;
for i=1:rows
for j=1:cols
C(X(m, 1), X(m, 2), :) = I(i, j, :);
m = m + 1;
end
end
This works but in the result there are many pixeles without values. I guess, when I do "X2 = X*M", the range of the image on the transformation it's not same of the source and many values lost

Answer 1

If you have the Image Processing Toolbox, I would just use imrotate to do the rotation for you.
out = imrotate(I, 45);
Otherwise, I would try to vectorize your approach and use interp2. You can rotate all of the pixel centers by the specified angle and then sample at these rotated point.
% Compute the coordinates of all pixel centers
[x, y] = meshgrid(1:size(I, 2), 1:size(I, 1));
% Compute the rotation matrix
R = [ cos(deg) sin(deg);
-sin(deg) cos(deg)];
xy = [x(:), y(:)];
% Compute the middle point
mid = mean(xy, 1);
% Subtract off the middle point
xy = bsxfun(#minus, xy, mid);
% Rotate all of these coordinates by the desired angle
xyrot = xy * R;
% Reshape the coordinate matrices
xy = reshape(xy, [size(x), 2]);
xyrot = reshape(xyrot, [size(x), 2]);
% Interpolate the image data at the rotated coordinates
out = interp2(xy(:,:,1), xy(:,:,2), I, xyrot(:,:,1), xyrot(:,:,2));