Physically-based painting style 3D image synthesis using GPU

Abstract

Realistic objects simulation is important in visual reality and its performance is impacted by physical shape computation. Due to fast development and robust programmability of graphics hardware recently, much research takes advantage of GPU (graphics processing unit) computation to accelerate simulations in virtual environment. This paper presents painting style synthesis that fully simulates image behaviors and synthesizes Chinese painting images on the programmable graphics hardware. Also, the physically-based approaches can provide impressive simulation results, emulate deformable objects just like the ones in real world. Nevertheless the computation involves several physical principles and requires large amount of computation resources. As a result, real time simulation is hard to achieve. Therefore, the ability to provide a deformable object simulation in painting style synthesis interactively or even in real-time is what we concern with. Furthermore, we propose two methods to accelerate ink simulation. Both methods are implemented using special functionalities on the modern graphics hardware. As result, our speedup implementation several times faster than original one.