標題: 以小波轉換壓縮並瀏覽漫畫文件
Compression and Browsing through Strip Comic with Wavelet Transform
作者: 宣志凌
Chih-Lin Hsuan
李素瑛
Suh-Yin Lee
資訊科學與工程研究所
關鍵字: 漫畫;影像切割;文字辨識;小波轉換;Comic;Image segmentation;OCR;Wavelet transform
公開日期: 1999
摘要: 在多媒體與網際網路盛行的今日,影像壓縮技術佔有非常重要的地位。如何以最小的資料量來儲存傳送影像是影像壓縮的課題。傳統的JPEG壓縮法適用於一般影像,但是當壓縮比增大時,影像品質相對降低。因此研究針對特定類型影像的新壓縮技術就因應而生。 本篇論文中,我們以小波轉換(wavelet transform)的影像分析壓縮技術為基礎,提出一個針對漫畫式圖文混合文件的壓縮法。因為人眼視覺對圖形跟文字所要求的品質不同,所以在壓縮時我們可以利用OCR文字辨識的技術分離出文字與圖形,然後再針對兩者做不同壓縮處理。最後依據使用者閱讀漫畫式文件的方向習慣來儲存各圖形區塊的小波轉換頻譜,如此可以讓使用者在閱覽時能快速地看到最重要的資訊,同時兼顧圖形品質及壓縮比率。 以本論文的方法,可以大大提升無線環境等低網路頻寬狀態下閱讀漫畫式文件的品質,同時此技術也可以應用至其他類似的圖文混合文件中。
With the growth of multimedia applications and the Internet, Image compression technique becomes more and more important. We concentrate on how to store and transmit multimedia data with minimum resources. The most common solution is to use JPEG compression technique. JPEG has excellent performance under normal condition, but with the increase of compression ratio, the quality of the image decays quickly. To solve this problem, we need new compression technique to deal with specific multimedia applications. In this thesis, we present a new solution for the digitalization of Comic documents using image encoding based on wavelet transform. By optical character recognition (OCR) technique and Hough transform line detection in image processing, the input comic document will be separated into dialogs and comic boxes. We re-order the wavelet frequency bands of each dialog and each comic box according to reading habit. By this method, Comic Documents can be transmitted more efficiently through limited network bandwidths and a more pleasant browsing environment is provided for readers, especially under wireless network and mini browser screen such as WAP applications.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT880392021
http://hdl.handle.net/11536/65417
顯示於類別:畢業論文