JF-Cut: A Parallel Graph Cut Approach for Large-scale Data

Graph Cut has proven to be an effective scheme to solve a wide variety of segmentation problems in graphics community. The main limitation of conventional graph-cut implementations is that they can hardly handle largescale datasets because of high computational complexity. Even though there are some parallelization solutions, they commonly suffer from the problems of low parallelism (on CPU) or low convergence rate (on GPU). In this paper, we present a novel graph cut algorithm that leverages a parallelized jump flooding (JF) technique and a heuristic pushing and relabeling scheme to enhance the components of the graph cut process, namely, multi-pass relabel, convergence detection and block-wise push-relabel. The entire process is parallelizable on GPU, and outperforms existing GPU-based implementations in terms of global convergence, information propagation, and performance. We design an intuitive user interface for specifying interested regions in cases of occlusions when handling volumetric data or video sequences. Experiments on a variety of datasets, including images (up to 15K × 10K), videos (up to 2.5K × 1.5K × 50), and volumetric data, achieve high quality results and a maximum 164-fold speedup over conventional approaches.