A Highly-Effective Approach for Generating Delaunay Mesh Models of RGB Color Images

Abstract

A highly effective method for generating Delaunay mesh models of RGB (i.e., red-green-blue) color images, known as CMG, is proposed. This method builds on ideas from the previously-proposed GPRFSED method of Adams for grayscale images to produce a method that can handle RGB color images. The key ideas embodied in our CMG method are Floyd-Steinberg error diffusion with improved initial-condition selection and greedy-point removal. Through experimental results, our CMG method is shown to outperform several competing methods that are based on a straightforward extension of grayscale mesh generators to color, with our method yielding meshes of vastly better quality at lower or comparable computational/memory cost.