視訊內容描述與索引

Abstract

多媒體的技術在今日已經十分成熟，並已經實際應用在許多不同的領域中。但相對的隨著多媒體的廣泛應用，其資料量也跟這日益增多，因此對多媒體的內容作分類，搜尋方法，多媒體檔案的管理等問題也越來越受到大家的重視。因此最近MPEG組織亦想發展一套針對多媒體內容作描述及管理的標準，那便是MPEG7產生的最主要的目的了。 今日的世界中，影片在我們的生活中到處可見，並且數位化的影片資料也蔚為風潮，因此數位資料的管理分類等問題也漸漸受到大家的重視。過去對於影片的管理大多是採用人工標記的方式，以此為影片的分類及搜尋的依據。藉由人工去分析該部影片的內容，將其依據內容性質作分類，更詳細的再將該部影片的內容由人工做整理，以文字簡單敘述該部影片的內容。 傳統方式的解決辦法應用在今日數位的影片資料將會面臨到許多問題。首先是處理大量影片資料時用人工方式來作分類，以及對內容的摘要描述，將無法應付日益增多的數位影片資料。再來是必須將所有的影片作分類，如果以人工的方式來操作也會有標準不一的問題產生。因此較好的解決方式就是應用現今的資訊技術，利用電腦來作分類及對內容摘要的描述。如此便可以有效的解決上述的問題。 因此這篇論文分為兩部分，首先第一部份是對影片作摘要,我們利用現今許多關於影片摘要擷取的技術,特別是在Motion Vector這方面的研究,並希望得到較有意義的影片內容的摘要.而關於影片的部分是採用ＭＰＥＧ-2為多媒體影片的資料來源，利用現今多媒體資訊的技術針對對一部電影內容作摘要影片的擷取，並在同時利用篩選出該摘要的重要特徵出來.第二部分再以XML的技術為原則產生Description Schemes (DSs)，再以先前所做好的影片摘要為基礎，做成對該影片內容的Descriptor。最後利用做好的Descriptor來對影片作階層分類及搜尋的功能,依此種方法所做出來的分類可能是就針對影片內容分類而言是較好的選擇。其後再利用簡單的介面，可讓使用者能透過網際網路，對該影片的內容做簡單的瀏覽，並可實際傳回完整的摘要影片做進一步的了解。其目的在實際結合自動摘要影片內容並在同時作對該影片內容相關的Descriptor，如此便可達到自動化的效果以及標準統一的目的。希望能對將來的數位圖書館，電影，新聞影片，等相關影片的搜尋及管理有所助益。

The recent rapid proliferation of multimedia content is increasing the need for efficient methods for searching, indexing, and managing multimedia. Therefore, content-based queries and indexing is more important in the future. The MPEG group recently established the MPEG-7 effort to standardize the multimedia content interface. The proposed interface will bridge the gap between various types of content meta-data, such as content features, annotations, relationships, and the search engines. Video is very popular today. Digital video is become the rising tide of multimedia, like MPEG-1 and MPEG-2. The amount of video data is increasing quickly. Thus indexing and cataloging of digital videos are more and more important. The best way for indexing and cataloging the video data is by its content. In the past, we most describe and annotate the content of video manually. But this traditional solution is not suitable for the enormous amount of video data. We must find a method that can generate the summary and description automatically, and provide an efficient and flexible method to illustrate the video content for users. In this paper, our system is divided into two parts. In the first part, we want to extract the summary of video content. We use many video summary technologies, special in the motion vector research, and want to get the more meaningful summary about the video content. We use MPEG-2 video stream to be our original video stream. We want to get more meaningful summary stream from the original video stream. At the same time, we will extract the important features of the summary, and use these features to represent this summary. The second part is that we use XML technology to generate extensible description schemes (DSs). Following this DSs, we can generate the descriptors (Ds) of the summary. When these descriptors are produced, we can use these descriptors to index and catalog these videos. Our indexing and cataloging results may be more reasonable and meaningful about the video content. We implement a system that can provide browsing and query functions for the users. Users can browse a video content from WWW and download these videos that they are interested. Our main purpose is to put summarizing and describing procedures together. Generating the summary and descriptors concurrently. Our solutions are closer to the principle of automatism and uniformity.