Video summarization based on key-places clustering and logical story units

基于关键地点聚类和逻辑故事单元的视频摘要

• 批准号: 227071-2011
• 负责人: Gagnon, Langis
• 金额: $ 1.6万
• 依托单位: Centre de recherche informatique de Montréal
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2015
• 资助国家: 加拿大
• 起止时间: 2015-01-01 至 2016-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=570234
• 关键词: Video; summarization; based; key; places

Video content can be accessed by retrieval or browsing. Retrieval is useful when the user know exactly what he/she is looking for. Browsing is useful when the user only needs an overview of the content. In a sense, retrieval is like looking for something from the index of a book while browsing is from the table of content (ToC). Video summarization is the process of generating an audio-visual ToC to help quickly capture the content of a video. The proposed research program targets video summarization. The objective is to develop and integrate multimodal analysis tools to segment a video into logical story units (LSU) based on a new automatic key-places clustering. This will be integrated into a LSU-based GUI, rather than a standard shot- or key-frame-based one, that provides a graphical ToC to the user with the possibility to access parts of the video from it. The GUI should be adaptable to allow user to access video based on his preferences and compatible with the TRECVID evlauation. Video indexing is a very active research topic. Many researches are still needed to reduce the so-called "semantic gap" between low-level features and human understanding. This project wants to be a contribution to this important field but through the video summarization "lense". In particular, a LSU-based videosummarization GUI will be an original contribution of the research program for the advancement of knowledge in the field of visual content usability and for the information technology sector.

可以通过检索或浏览来访问视频内容。当用户确切地知道他/她在寻找什么时，检索是有用的。当用户只需要内容的概览时，浏览是有用的。从某种意义上说，检索就像是从书的索引中查找内容，而浏览则是从内容表（ToC）中查找。视频摘要是生成视听ToC以帮助快速捕获视频内容的过程。 所提出的研究计划的目标视频摘要。目标是发展和整合 多模态分析工具，将视频分割成逻辑故事单元（LSU）， 关键点聚类这将集成到基于LSU的GUI中，而不是标准的快照或 基于关键帧的一种，向用户提供图形ToC，用户可以从中访问视频的部分内容。GUI应可自适应，允许用户根据其偏好访问视频，并与TRECVID评估兼容。 视频索引是一个非常活跃的研究课题。还需要进行许多研究来减少所谓的 低层特征与人类理解之间的“语义鸿沟”。该项目希望通过视频摘要“镜头”对这一重要领域做出贡献。特别是，一个基于LSU的视频摘要GUI将是一个原始的贡献的研究计划，在视觉内容的可用性和信息技术部门的知识进步。