跳到主要內容區塊

 

99 有關Flash 動畫中影像蒐集技術的探討

出版年度:99

 

論文名稱:有關Flash 動畫中影像蒐集技術的探討

 

研究生:陳綮紳   指導教授:游耿能

 


論文簡介:

  在影像搜尋引擎的研究領域中,通常會依賴網路爬行器(Web Crawler)去各網 站蒐集網頁上的影像與文字。不過根據目前的HTML 規格書,網頁上利用IMG 元 素所能呈現的影像格式僅有GIF、JPEG 和PNG 三種,而Flash 動畫除了涵蓋上述 三種影像格式之外,更包括了PIC、TIFF、BMP 等影像格式。雖然Flash 動畫可涵 蓋的影像格式甚多,但針對影像格式的儲存方式主要還是分為失真與無失真影像 二大類。由於絕大部份現有的網頁爬行器鮮少針對Flash 動畫中的影像進行萃取並 蒐集,所以本研究設計實作一個Flash 動畫蒐集器Swiler,不僅利用Swiler 在網路 上所蒐集到的Flash 動畫,進行判斷並分析是否內嵌影像資料,並將這些影像資料 分別地萃取並輸出對應的影像格式,再加上Flash 動畫檔可能從外部來載入各種影 像,Swiler 也一併地俱備加以蒐集的功能。從實測的結果中,Swiler 不僅能達到對 Flash 動畫及其內部的影像蒐集,也改善影像搜尋引擎,增加蒐集的影像數量,未 來將可進一步地針對檢索Flash 動畫的相關議題進行研究。In the design of image search engine related domain, web crawlers usually navigate the content of web pages to collect images and text. According to the HTML specification, the IMG tag can only employ three compressed image file formats, namely, GIF, JPEG and PNG, to describe images for web pages. In addition to these three image format, the flash movies’ SWF file may also incorporate several other image formats such as PIC, TIFF and BMP. Though flash movies utilize these versatile image formats, these image formats can be divided into two categories, namely, lossless and lossy images. Most current web crawlers are unable to collect or extract these embedded images from flash movies. Therefore, we try to harvest images from Flash movies on the internet in this study. We explore the harvesting techniques and propose a system prototype called Swiler with a novel crawling technique to facilitate the harvesting. We develop algorithms to extract Flash movies on the internet first. Then we analyze the format of the embedded images within flash movies and develop the image harvester with novel extracting method. And finally, we develop algorithms with threading technique to enhance the performance of harvesting. We also take the harvesting of images external to the flash movies into account by augmenting the Swiler crawler with path-crawling features to collect these images. In our preliminary experimental tests of the swiler, the improvement of our approach is reflected in the increased number and the extensiveness of extracted images and flash movies. Future work might focus on various issues of enhancing the retrieval or the utilization of these Flash movies and images.

瀏覽數