對於以內容為主的多媒體資訊索引與摘要的需求愈來愈高時，擷取出有內涵意義的特徵值變成為一份重要的課題。於數位視訊中，畫面的文字即是十分有用的特徵值，它不僅可以清楚表達出該影片的內涵，而且並不難以擷取。再者，相較於語音辨識或是視覺影像分析的不完善，文字辨識系統卻已趨近成熟而完整。因此，大多數的視訊索引系統研究一開始以文字辨識為濫觴。 在此篇論文，我們提出針對於移動文字之偵測與擷取演算法。相較於固定字幕的演算法而言，少有研究針對於移動文字。我們先利用Sobel detector找出可能為文字邊緣的像素，再使用垂直與水平統計表定位出正確的文字區域，最後採用Otsu Method 決定出臨界值以區分出文字與背景。不幸地，此方法仍有少數非文字的像素被辨識為文字。在此，我們使用提出的modified seed-fill演算法消除錯誤辨識的非文字區塊以提升辨識率。根據實驗結果，所提出的演算法對於不同類型視訊都能提供不錯的結果。 Text in video is a very compact and accurate clue for video indexing and summarization. Most video text detection and extraction methods deal with the static videotext on video frames. Few methods can handle motion videotext well since motion videotext may hardly be extracted well. In this thesis, we propose a low computation load text detection and localization method to detect and localize the scrolling videotexts which provide much information for us. We also propose a videotext extraction method to extract the videotext. The detection method is carried out by edge detection, and the projection profile method is used to localize the text region well. The extraction method consists of adaptive thresholding, and our proposed modified seed-fill algorithm. Experimental results on a large number of video images are reported in detail.
