Image Metamorphosis based on Elastic Body Spline and Snake Model

Image metamorphosis ( For short, image morphing ) is an active eld of research and recent e orts aim at improving both user interface and warping results. Generally, the morphing technique involves three problems. The rst is to establish correspondence between given two images, which is most di cult part of morphing process. The correspondence is usually established by hand. The second problem is to de ne or construct a mapping function which deforms the rst image towards the second one based on these feature points, and vice versa. The third problem is to blend pixel values of the two deformed images to create in-between images, this will end the morphing process. This study aims to raise strategy to solve these problems. First, we propose a semi-automatic algorithm based on snake model to specify the feature correspondence between two given images. It allows a user to extract contour that de ne facial feature such as lips, mouth, pro le, etc., by specifying only endpoints of the contour around the feature that serves as the extremities of contour. Then we use these two points as anchor points, and automatically computes the image information around these endpoints to provide boundary conditions. Next, we optimize the contour by taking this information into account close to its extremities. During the iterative optimization process, the image forces are moving progressively from the contours extremities towards its center to de ne the feature. It helps the user to de ne easily the exact position of a feature. It may also reduce the time taken to establish feature correspondence between two images. For the second image morphing problem, this paper presents a new image warping algorithm which uses elastic body spline to generate a warp function that interpolates scattered data points. The spline is based on a Partial Di erential Equation of Naiver that describes the equilibrium displacement of the elastic body subjected to forces. It generates a smooth warp that re ects the feature correspondence points and has several advantages. It is e cient in smoothed interpolated morphed images with only a remarkably small number of feature points speci ed. Finally, we describe an e cient cross dissolve function to generate the in-between images. The in-between images could be de ned by creating a new set of feature points through the cross-dissolving process.