Emilie Dumont

Split-screen dynamically accelerated video summaries

In this paper, we describe our approach to the TRECVID 2007 BBC Rushes Summarization task. Our processing is composed of several steps. First the video is segmented into shots. Then, one-second video segments are clustered into similarity classes. The most important non-redundant shots are selected such that they maximize the coverage of those similarity classes. Then shots are dynamically accelerated according to their motion activity to maximize the content per time unit. Finally they are optimally grouped by sets of four to be presented using split-screen display. The summaries produced have been evaluated in the TRECVID campaign. We present a first attempt at automating the evaluation process.

Automatic Story Segmentation for TV News Video Using Multiple Modalities

While video content is often stored in rather large files or broadcasted in continuous streams, users are often interested in retrieving only a particular passage on a topic of interest to them. It is, therefore, necessary to split video documents or streams into shorter segments corresponding to appropriate retrieval units. We propose here a method for the automatic segmentation of TV news videos into stories. A-multiple-descriptor based segmentation approach is proposed. The selected multimodal features are complementary and give good insights about story boundaries. Once extracted, these features are expanded with a local temporal context and combined by an early fusion process. The story boundaries are then predicted using machine learning techniques. We investigate the system by experiments conducted using TRECVID 2003 data and protocol of the story boundary detection task, and we show that the proposed approach outperforms the state-of-the-art methods while requiring a very small amount of manual annotation.

Rushes video summarization and evaluation

During the post-production stage of film making, the film editor is faced with large amounts of unedited raw material, called rushes. Developing tools to view and organize this material is an important component of video processing. This paper describes an approach for summarizing rushes video based on the detection of repetitive sequences, using a variant of the Smith-Waterman algorithm to find matching subsequences. We rely on the evaluation methodology that has been introduced in the TRECVID BBC Rushes Summarization Task. We propose an automation of the manual TRECVID evaluation using machine learning techniques to train an automatic assessor. We compare the automatic assessor evaluation to the evaluations provided by the TRECVID manual assessors.

Rushes Video Parsing Using Video Sequence Alignment

In this paper, we propose a novel method inspired by the bio-informatics domain to parse a rushes video into scenes and takes. The Smith-Waterman algorithm provides an efficient way to compare sequences by comparing segments of all possible lengths and optimizing the similarity measure. We propose to adapt this method in order to detect repetitive sequences in rushes video. Based on the alignments found, we can parse the video into scenes and takes. By comparing takes together, we can select the most complete take in each scene. This method is evaluated on several rushes videos from the TRECVID BBC Rushes Summarization campaign.

Automatic evaluation method for rushes summary content

During the last years, the developpment of rushes video summarization systems has greatly increased thanks to the international evaluation campaign TRECVID. In this paper, we propose an automation of the manual TRECVID evaluation using machine learning techniques. We train an automatic assessor to perform evaluation on summary content and we show a high correlation between the manual evaluation performed in TRECVID 2008 and our automatic assessor.

Sequence alignment for redundancy removal in video rushes summarization

In this paper, we describe our approach to the TRECVID 2008 BBC Rushes Summarization task. First, we remove junk frames and dynamically accelerate videos according to their motion activity to maximize the content per time unit. Then, we search identical sequences using a sequence alignment algorithm derived from bio-informatics and we identify and structure scenes in videos, then we select one take per scene. We select the most relevant sequences in order to maximize the content and finally, we compose our summary in an original presentation. The produced summaries have been evaluated in the TRECVID campaign.

Rushes video summarization using a collaborative approach

This paper describes the video summarization system developed by the partners of the K-Space European Network of Excellence for the TRECVID 2008 BBC rushes summarization evaluation. We propose an original method based on individual content segmentation and selection tools in a collaborative system. Our system is organized in several steps. First, we segment the video, secondly we identify relevant and redundant segments, and finally, we select a subset of segments to concatenate and build the final summary with video acceleration incorporated. We analyze the performance of our system through the TRECVID evaluation.

A collaborative approach to automatic rushes video summarization

Video summarization is a useful tool which allows a user to grasp rapidly the essence of a video. In the development of this research topic we propose a new method based on different individual content segmentation and selection tools in a collaborative system. The main innovation of this work is to merge results from different approaches, so as to benefit from their respective qualities. Our system is organized in two phases: first segmentation of the video, second identification of relevant and redundant segments. The final list of selected segments is used to concatenate the video segments and build the final summary. In order to assess the effectiveness of this organization, we evaluate our system with a method based on the TRECVID 2007 BBC rushes summarization evaluation pilot and compare our performance with existing systems.

Automatic evaluation method for rushes summarization: Experimentation and analysis

Evaluation remains an important difficulty in the development of video summarization systems. Rigorous evaluation of video summaries generated by automatic systems is a complicated process because the ground truth is often difficult to define, and even when it exists, it is difficult to match with the obtain results. The TRECVID BBC evaluation campaign has recently introduced a rushes summarization task and has defined a manual evaluation methodology. In this paper, we explore the use of machine learning techniques to automate this evaluation. We present our approach and describe the current results, in comparison with manual evaluations performed in the 2007 campaign.

A Local Temporal Context-Based Approach for TV News Story Segmentation

Users are often interested in retrieving only a particular passage on a topic of interest to them. It is therefore necessary to split videos into shorter segments corresponding to appropriate retrieval units. We propose here a method based on a local temporal context for the segmentation of TV news videos into stories. First, we extract multiple descriptors which are complementary and give good insights about story boundaries. Once extracted, these descriptors are expanded with a local temporal context and combined by an early fusion process. The story boundaries are then predicted using machine learning techniques. We investigate the system by experiments conducted using TRECVID 2003 data and protocol of the story boundary detection task and we show that the extension of multimodal descriptors by a local temporal context approach improves results and our method outperforms the state of the art.