Fast Shape-Simplifying Image Abstraction Using Graphics Hardware

This paper presents a new GPU-based method for creating abstracted representations of photographs. Based on the constrained mean curvature flow and the Shock filter, our approach simplifies both shapes and colors simultaneously while preserving and conveying the directionality of important features and shape boundaries. The level of abstraction can be intuitively controlled by iteratively and incrementally applying the algorithm. Note that the whole pipeline design is highly parallel, enabling a GPU-based implementation. Our GPU-based method outperforms the CPU-based one with two magnitudes of speedup. Several experimental examples are shown to demonstrate both the effectiveness and efficiency of the proposed method.