Streaming Graph-Based Hierarchical Video Segmentation by a Simple Label Propagation

In this paper, we present an approach to streaming graph-based hierarchical video segmentation by simple label propagation. Here, we transform the streaming video segmentation into a graph partitioning problem in which each part corresponds to one region of the video, furthermore, we apply a simple method for merging the segmentations of two consecutive blocks to achieve the temporal coherence. The spatial-temporal coherence is given, only, by color information instead of more complex features. We provide an extensive comparative analysis among our method and methods in the literature showing efficiency, ease of use, and temporal coherence of ours. According to the experiments, our method produces good results when applied to video segmentation besides presenting a low space and time cost, compared to other methods.