Digital Image Deformation

Image metamorphosis, a visual technique of transforming smoothly one digital image into another, is one kind of keyframe animation technique. In recent years, it has found widespread uses in entertainment, advertisement and education.Image metamorphosis technique basically deals with three problems: specification of image features, derivation of warp functions, and design of transition functions. Of these three, the most important one is how to derive the warp functons.The process of warping an image can be devided into two steps: spatial transformation and image resampling.Spatial transformation computes the desired displacements of all pixels in the source and destination images. The control elements( points or line segments discribing image features) used in metamorphosis are usually unregularly distributed, thus the spatial transformation there can generally be defined as scattered data interpolation.Resampling refers to the transformation of a sampled image from one coordinate system into another, and the two coordinate systems are related to each other by the mapping function of spatial transformation. The main problem of resampling is the selection of interpolating kernal.After comparing and contrasting several typical morphing methods, we present here a new algorithm -- metamorphosis based on star polygon segmentaion of images. The basic idea is to segment the source and destination images parallelly with star polygons according to the image features. This method can preferably specify image features, give good results, and make some improvements in speed and control.