Abstract
Application of geometric transformation to images requires an interpolation step. When applied to image rotation, the presently most efficient GPU implementation for the cubic spline image interpolation still costs about eight times as much as linear interpolation. The implementation involves two steps: a prefilter step performs a two-pass forward-backward recursive filter, then a cubic polynomial interpolation step is implemented thanks to a cascade of linear interpolations. This article proposes a simpler and faster implementation for the prefilter—which is the most time consuming—in terms of a direct convolution. The overall cost for our cubic B-spline interpolation algorithm then reduces to 4.5 times the cost of linear interpolation.
Acknowledgments
Aurélien Plyer is gratefully acknowledged for the many discussions about this work. Moreover, the authors salute the helpful comments of the referees.