Feature-based image metamorphosis optimization algorithm

Image metamorphosis using simple features provides flexible and easy control of visual effects. The feature-based image metamorphosis algorithm proposed by Beier and Neely (1992) suffers from two problems; speed and strange effects. The paper presents an optimization technique to recover these problems in a reasonable time. The idea is based on local decomposition of metamorphosis into piecewise approximation and then applying the above technique to establish one-to-one correspondence to recover such effects. The combination of all mappings produces a foldover-free result as well as dominating the ghost effect. The results show that our implementation is efficient in time complexity, smoothly interpolates metamorphosis images with only a remarkably small number of specified feature points, and the speed of computations makes image metamorphosis animation reliable for interactive machines.