Physics-preserving fluid reconstruction from monocular video coupling with SFS and SPH

Abstract

We propose a joint method to reconstruct dynamic fluid volume sequences from a monocular video. Compared with previous methods, sophisticated equipment or careful experimental setups are not required. In order to recover the surface detail and maintain its physical property, a joint reconstruction method coupled with shape from shading (SFS) and smoothed particle hydrodynamics (SPH) algorithms is proposed. SFS is the first used to recover the height field for each frame of the video, and then the height field of the first frame is further extended to a volume as the initial state of the SPH simulation. Based on the above initial data, the key idea of our method is to optimize a SPH model to simulate the fluid volume sequences conforming to the laws of physical motion, and correct the fluid volumes to refine the surface details conforming to the recovered height field by SFS. Our experimental results compare favorably to the state of the art in terms of global motion features and fluid surface details and demonstrate the performance of our approach.