Seamless NPR video painting

Abstract

The paper presents a method for automatic transformation of ordinary videos into non-photorealistic (NPR) graphical representations, for achieving painterly effects. The primary goal is the non-interactive transformation of videos, with the intended use as a service in community video databases (as an extension to content based retrieval). There are no restrictions on the types of videos. The main steps involve scene separation, color segmentation, foreground and focus area extraction, and vector graphical transformation of selected frames.