Multi-view Video Summarization using Bipartite match constrained Optimum-Path Forest Clustering

Multi-View Video Summarization using Bipartite Matching Constrained Optimum-Path Forest Clustering

Abstract :-

The task of Multi-View Video Summarization is to efficiently represent the most significant information from a set of videos captured for a certain period of time by multiple cameras. The problem is difficult to solve because of the huge size of the data, presence of many unimportant frames with low activity, inter-view dependencies and significant variations in illumination. In this paper, we propose a novel solution to the above problem within a graph-theoretic framework. Semantic feature in form of visual bag of words as well as visual features like color, texture and shape are used to represent each video frame. Gaussian entropy is then applied to filter out frames with low activity. Inter-view dependencies are captured via bipartite matching and optimum-path forest algorithm is applied for clustering of the video frames. Both subjective and objective evaluations clearly indicate the effectiveness of the proposed method.

Results

Multi-View Video Summarization using Bipartite Matching Constrained Optimum-Path Forest Clustering

We evaluate five Multi-view videos namely Office1, Campus, OfficeLobby, Road and Badminton, capturing both indoor and outdoor environments . All the five video datasets are typical surveillance videos, consisting of three to four views. Detailed information regarding these Multi-View videos and ground truths of videos were prepared and made available at http://cs.nju.edu.cn/ywguo/summarization.html We use these available ground truths of videos to validate the performance of our method.

1.Comparison with Standard K-means clustering Algorithm.

First we use a simple base method where K-means clustering algorithm is applied to visual features (color, edge and shape) to obtain the Multi-view summary. We choose K-means because of its low computational overhead in clustering of high dimensional data.

Result of Multi-view summary of given datasets by using k-means clustering algorithm are given as [Office1], [Campus],[OfficeLobby].

2.Our Purposed Method.

In this paper, we proposed a novel framework for the Multi-view video summarization using bipartite matching constrained OPF. The problem of capturing the inherent correlation between the multiple views was modeled as a MCMW matching problem in bipartite graphs. We introduced the OPF clustering for summarizing the multi-view videos with very effective results.

Result of Multi-view video summary of given datasets by using our proposed method are given as : [Office1] [Campus] [Office Lobby] [Road] [Badminton].