A rule based streaming video segmentation using minimum spanning tree algorithm

Abstract

Video segmentation is to partition a video into volumetric portions. It is relevant to a wide assortment of vision applications, such as activity acknowledgment, scene arrangement, video outline, content-based video recovery, and 3D remaking. The main objective of this method is to extract significant entities from a video. This video segmentation is classified into offline and online video segmentation algorithms. Compared to online video segmentation, offline video segmentation requires huge memory space. Nevertheless, the main challenge for video segmentation is to segment the online streaming video. However, one of the efficient streaming video segmentation algorithms is hierarchical segmentation algorithm. In this paper, an efficient Segmentation Approach for Streaming Videos Using a Rule with the Minimum Spanning Tree Algorithm is presented for further reducing the runtime and memory consumption. In this approach, each frame from the streaming video is modeled as a weighted graph. From the weighted graph, the minimum weighted hierarchical tree is formed by applying Kruskal's minimum spanning tree algorithm. Then the hierarchical segmentation is done by removing the inconsistent edges based on a rule. Performance of our proposed approach is compared with the existing work in terms of Boundary Precision-Recall (BPR) and Volume Precision-Recall (VPR).