When rotating a shape in context of rendering shapes with pixels, is it most efficient to rotate the shape and apply a sampling technique at each degree of rotation? Or is there another way to move this shape through the screen without having to resample?
anon
My intuition is that when rotating the image by an angle that does not maintain the same shape (ie. not 90 / 180 / 270 etc.), this will require resampling as the diagonal edges will no longer be perfectly straight edges.
wilsonlee
Can you expand on your response a bit? Why would straight edges no longer be straight when rotating? I thought rotations always preserve the original shape and size?
unicorn
I agree with @wilsonlee's response, and I am also confused by @anon's response to the original question.
In response to @et17's question, I would imagine that as a shape is rotated, it is only sampled at the refresh rate of the display rather than at every degree. It can move more than one degree per refresh, so sampling at each change in degree.
anon
Let's consider the two shapes represented in the slide.
In the first shape, during the rasterization process the edges of the square will be able to be perfectly "straight" edges in pixels.
In the second shape, during the rasterization process the edges of the square can no longer be represented as perfectly "straight" edges in pixels and this will result in jaggies.
This is what I meant by the diagonal edges no longer being straight edges. It seems to me like this means that resampling must take place in order to rotate the first shape to match the second shape.
When rotating a shape in context of rendering shapes with pixels, is it most efficient to rotate the shape and apply a sampling technique at each degree of rotation? Or is there another way to move this shape through the screen without having to resample?
My intuition is that when rotating the image by an angle that does not maintain the same shape (ie. not 90 / 180 / 270 etc.), this will require resampling as the diagonal edges will no longer be perfectly straight edges.
Can you expand on your response a bit? Why would straight edges no longer be straight when rotating? I thought rotations always preserve the original shape and size?
I agree with @wilsonlee's response, and I am also confused by @anon's response to the original question.
In response to @et17's question, I would imagine that as a shape is rotated, it is only sampled at the refresh rate of the display rather than at every degree. It can move more than one degree per refresh, so sampling at each change in degree.
Let's consider the two shapes represented in the slide.
In the first shape, during the rasterization process the edges of the square will be able to be perfectly "straight" edges in pixels.
In the second shape, during the rasterization process the edges of the square can no longer be represented as perfectly "straight" edges in pixels and this will result in jaggies.
This is what I meant by the diagonal edges no longer being straight edges. It seems to me like this means that resampling must take place in order to rotate the first shape to match the second shape.