High-fidelity image warping for serial and parallel processing

High-fidelity image warping preserves image intensity information accurately across a warp. It is important if warping is to be used transparently within general image processing applications, particularly when these involve long sequences of multiple warps such as in morphing. To achieve this aim we require near-perfect interpolation and the avoidance of aliasing. For fast parallel implementation we make use of multipass (scan-line) algorithms, which introduce additional problems. We refer to our previous work on FFT interpolation and ways to avoid aliasing in multi-pass rotation which help it to be a preferred choice for high fidelity work. We show how to visualise aliasing in warping, in practice, for both single-stage and multipass methods, using an artificial test image formed as a “spider web” pattern in the Fourier domain