Suppose I took a picture of a whiteboard at an angle. If I could detect where the corners of the whiteboard are, what algorithms and linear projections could I use to convert this slanted view into a "head on" view?
$\begingroup$
$\endgroup$
-
$\begingroup$ Check out the answer to this question. You can take any 4 points and transform the quadrilateral between them into a unit square. From there you can simply scale in x and y to the size and shape you want. $\endgroup$ – user1118321 Oct 26 '17 at 4:57
-
$\begingroup$ Does the transformation need to be perspective correct ? $\endgroup$ – PaulHK Oct 26 '17 at 12:27
-
$\begingroup$ @user1118321 thanks for the link, that's perfect! $\endgroup$ – Lincoln Bergeson Oct 26 '17 at 15:38
-
$\begingroup$ @PaulHK what do you mean by perspective correct? $\endgroup$ – Lincoln Bergeson Oct 26 '17 at 15:38
-
$\begingroup$ For example, if the floor was a huge chess board, we would see squares vanish into the distance. When transforming back to 2d we need to take that perspective into account. $\endgroup$ – PaulHK Oct 27 '17 at 2:40
$\begingroup$
$\endgroup$
As @user1118321 mentioned, as well as @NathanReed in this answer to another question, you can do a perspective transform to transform any convex quadrilateral to any other.
OpenCV has a good library for doing perspective transformations if you don't want to roll your own.
