Suppose I have some arbitrary input image with width w1 and height h1. I want to rotate this image all the way around 360 degrees back to the starting position. However, if the image is of anything except a circle, then the edges of the image will be clipped if the canvas it is drawn onto remains at size w1 by h1.
What I need then is to determine the canvas size (width w2 and height h2) that can be used for all rotated versions of the input image. I know that w2 == h2, and thus the desired canvas size is a square, because it is obvious that we are rotating something about a center point, and the final image (after 360 rotations) is essentially a circle.
The other thing to consider is that objects such as squares will have corners that will stick out, so just using the max value of width or height for both dimensions also does not work.
One solution I have come up with is to create the canvas larger than I need (for example by setting w2 and h2 to max(w1, h1) * 2, rotating everything, and then trimming all the transparent pixels. This is not very efficient and I would much rather be able to calculate the tightest bounding box upfront.
