Summary: AN EVALUATION OF SEVERAL MESH-GENERATION METHODS USING A SIMPLE MESH-BASED IMAGE CODER
Michael D. Adams
Department of Electrical and Computer Engineering, University of Victoria, Victoria, BC, V8W 3P6, Canada
THE IMAGE REPRESENTATIONS used by most conventional image
coders employ regular (e.g., lattice) sampling. Due to the nonsta-
tionary nature of most images, however, such sampling is usually highly
inefficient. Inevitably, when regular sampling is employed, the sampling
density will be too low in regions where the signal is rapidly changing,
and too high in regions where the signal is varying slowly or not at all.
The above problem can be overcome by employing an image repre-
sentation that facilitates the use of arbitrary (i.e., irregular) sampling.
In this context, image representations based on triangle meshes are of
particular interest, as such meshes are ideally suited to accommodat-
ing arbitrary sampling. One of the major challenges associated with
mesh-based representations is finding effective methods for construct-
ing a mesh that accurately represents a given image (i.e., the so-called