A parallel one dimensional resampling algorithm

Abstract

Separable algorithms are useful in casting image warping into a framework that is amenable to hardware implementation. Growing interest in this area has gained impetus from the widespread proliferation of advanced workstations and digital signal processors. Examples include realtime hardware for video effects among others. The central benefit of separable algorithms is the reduction in complexity of one dimensional resampling algorithms. When the input is restricted to be one-dimensional, efficient solutions are made possible for the image reconstruction and antialiasing components of resampling. Fant (1986) presented such a solution that is well suited for hardware implementation. Fant's algorithm, however, is inherently serial. We present an alternative but equivalent (in the sense of producing exactly the same output scan lines) algorithm, that is parallel. The new algorithm was inspired by Wolberg's algorithm (1990). The new algorithm adds some overhead to the hardware implementation of Fant's algorithm to be able to achieve parallelism.