Features of interpolations algorithms for resizing
Posted: Wed Jun 30, 2004 10:27 am
First, thanks for adding more well-proofed interpolation kernels into XnView.
Practical Question:
Are the new interpolations only used in resizing or are they also used in all other affine transformations like rotation ...?
Background Questions:
What kernel sizes for the higher interpolation is XnView using? I suspect from first tests that you did not choose very big kernels. I love to have kernels smaller than 4x4 since they are faster and since the infamous, degrading Gibbs rings are still quite small.
What is the prefiltering in the interpolation for enlarging and what is it for shrinking images? I do not see any aliasing when shrinking images, so there must be filters in the interpolation. Does it degrade the quality so much?
Practical Question:
Are the new interpolations only used in resizing or are they also used in all other affine transformations like rotation ...?
Background Questions:
What kernel sizes for the higher interpolation is XnView using? I suspect from first tests that you did not choose very big kernels. I love to have kernels smaller than 4x4 since they are faster and since the infamous, degrading Gibbs rings are still quite small.
What is the prefiltering in the interpolation for enlarging and what is it for shrinking images? I do not see any aliasing when shrinking images, so there must be filters in the interpolation. Does it degrade the quality so much?