Speedup techniques for text image compression with JBIG2
Y. Ye, Pamela C. Cosman · 2001
Pattern matching is the most time consuming process in text image compression with JBIG2. In this paper we propose three techniques to speed up the pattern matching process. By limiting the search range for matching symbols in the dictionary, the first technique saves 15% of encoding time with almost no bit rate penalty By making early decisions about the pattern matching outcome, the second technique saves another 15% of encoding time with no coding loss. The third technique looks at enhanced prescreening using additional symbol features besides symbol size. Using certain topological features, enhanced prescreening can save up to 75% of encoding time with at most 1.6% of bit rate penalty.