0

I have an image full_image_map which I have rotated by a certain angle using the following code: 

Mat rotated_full_map_image;
Point2f src_center(full_map_image.cols/2.0F, full_map_image.rows/2.0F);    
Mat rot_mat = getRotationMatrix2D(src_center, angle, 1.0);
warpAffine(full_map_image, rotated_full_map_image, rot_mat, full_map_image.size());

I have also found the coordinates of a point in the rotated image using this: 

Point rotated_centre;
Point2f p(some_x,some_y);  
rotated_centre.x = rot_mat.at<double>(0,0)*p.x + rot_mat.at<double>(0,1)*p.y + rot_mat.at<double>(0,2);
rotated_centre.y = rot_mat.at<double>(1,0)*p.x + rot_mat.at<double>(1,1)*p.y + rot_mat.at<double>(1,2);

However, my original image full_image_map is much more tall than it is wide. So when I rotate it by say 90 degrees and specify the size of the rotated image to be full_map_image.size(), most of the rotated image is cut off since it tries to fit it only in that mentioned size. This is similar to this answer.

My question: how can I properly perform this rotation by any given angle and still use the same method to find the coordinates of a point in the rotated image? I am using opencv with c++.

RaviTej310
  • 1,635
  • 6
  • 25
  • 51
  • You should be able to use the [accepted answer](https://stackoverflow.com/a/24352524/4228275) of the [question you linked](https://stackoverflow.com/questions/22041699/rotate-an-image-without-cropping-in-opencv-in-c): if you multiply the point with the rotation_matrix `rot` you should be fine. – Berriel Feb 16 '19 at 14:10
  • if you want a solution for any homography transformation (not only rotation), create a vector with all 4 corners of the original image, transform them with the homography, compute bounding box and adjust the homography according to the bbox position. – Micka Feb 16 '19 at 18:07

0 Answers0