Comaparing image-processing operators by means of visible differences predictor

Abstract

Utilization of non-photorealistic techniques (NPR) is beneficial in many cases, in comparison with the use of tradi-tional rendering methods in computer graphics. The observer's sensation is often straighter, clearer, or even more valuable. There exists plenty of various NPR techniques in computer graphics, however application of one tech-nique to the specific problem is not necessarily as providential as usage of another one. There arises a strong need to classify NPR techniques with respect to their applicability, to find a mechanism able to compare NPR techniques automatically. The search for such a mechanism will be a long term goal. We present our first steps towards the solution to this problem — the comparison of 2D-based NPR techniques using Daly's Visible Differences Predictor (VDP). Results from our experiments are reported and will be used to improve communication in graphical user interfaces.