Carleton University - School of Computer Science Honours Project
Fall 2019
Automated Stylized Reconstruction of Images using Luminance
Megan Perera
SCS Honours Project Image
ABSTRACT
The problem at hand is to recreate the subject of an image by constructing a mesh for the subject using triangles and filling the triangles such that each triangle is filled with lines parallel to one of its edges, and the full mesh is filled in a way which captures and reflects the original image in both content and lighting. To accomplish this, we use Canny edge detection to find the edges of the image, Poisson distribution to sample internal points that are used for mesh generation, use Delaunay triangulation to generate a mesh with the selected points, and find the luminosity range within the image so that each triangle we construct will be the correct shade. By utilizing these techniques, we are able to generate abstract images that respect the original image yet maintain a distinct artistic style.