ID-Based Rendering of Silhouettes on GPU

Abstract

When rendering object-silhouettes preprocessing is generally done primarily on the CPU. To this end primitive normals must be made consistent and the silhouette-edges need to be extracted every time the view-point is changed. In this paper we propose a pure image-based GPU-method where IDs of triangles are rendered to a texture and silhouettes are extracted based on the information stored in that texture. With this method the geometry does not need to be preprocessed or reprocessed when the view-point or the geometry is changed. Another important advantage of the proposed method over any Z-buffer based method is that it does not require any threshold value to compare against the difference between depth-values of the neighboring pixels which is difficult to adjust in perspective projection.