Abstract
In this work, we propose a novel approach to videosegmentation that operates in bilateral space. We design anew energy on the vertices of a regularly sampled spatio-temporal bilateral grid, which can be solved efficiently us-ing a standard graph cut label assignment. Using a bi-lateral formulation, the energy that we minimize implic-itly approximates long-range, spatio-temporal connectionsbetween pixels while still containing only a small numberof variables and only local graph edges. We compare toa number of recent methods, and show that our approach achieves state-of-the-art results on multiple benchmarks in a fraction of the runtime. Furthermore, our method scales linearly with image size, allowing for interactive feedback on real-world high resolution video.