3d Morphing - 6.837 Final Project Proposal

This paper presents our work towards achieving a model based approach to three dimensional morphing. It describes the initial algorithms and ideas that we envisioned, the final algorithm we developed and implemented, the environment we worked in, our visualization techniques, and future work planned on the subject. Introduction: Different types of morphing Morphing is an interpolation technique used to create from two objects a series of intermediate objects that change continuously to make a smooth transition from the source to the target. Morphing has been done in two dimensions by varying the values of the pixels of one image to make a different image, or in three dimensions by varying the values of three-dimensional pixels. We're presenting here a new type of morphing, which transforms the geometry of three dimensional models, creating intermediate objects which are all clearly defined three-dimensional objects, which can be translated, rotated, scaled, zoomed-into. Two-dimensional morphing Two-dimensional morphing is transforming an array of m by n pixels into another array progressively. An intermediate value between two pixels can be obtained by interpolating rgb values of the source and end pixels in more or less complicated ways. However, straight color interpolation creates many unwanted side effects such as ghosting and unnatural transitions. A better way to accomplish two-dimensional morphing is to identify line segments on the source image with line segments on the target image so that pixel values will actually move across the image so that features will be preserved better. For example to map a face to another face, it is important that certain features such as the eyes, nose, and mouth are identified so that intermediate images actually look natural. The mouth of the source image will move to the proper place in the target image. Figure 1: Two dimensional morphing of images Prior three-dimensional morphing Three dimensional morphing has been done using more or less the same technique. Instead of dealing with pixels in a two-dimensional image though, the people who did this used pixels in a three dimensional structure. The algorithms however are still the same, the features identified being now points, edges, cubes, and other three-dimensional structures.