Compound Image Compression with Multi-step Text Extraction Method

Texts appearing on compound image are usually classified into scene text and imposed text. Imposed text, like scripts in videos, slogans in advertisements and titles in magazine covers, contains important information. In this paper, a multistep compression method that could preserve imposed text quality in restored image is presented. The method separates imposed text from background and compresses the two with different schemes. Experiment results show satisfying text-image separation could be achieved, and clear texts could be restored under high compression ratio.

[1]  Michael R. Lyu,et al.  A robust statistic method for classifying color polarity of video text , 2003, 2003 International Conference on Multimedia and Expo. ICME '03. Proceedings (Cat. No.03TH8698).

[2]  Bing-Fei Wu,et al.  Algorithms for compressing compound document images with large text/background overlap , 2004 .

[3]  Wen Gao,et al.  Multi-polarity text segmentation using graph theory , 2008, 2008 15th IEEE International Conference on Image Processing.

[4]  Edward K. Wong,et al.  A new robust algorithm for video text extraction , 2003, Pattern Recognit..