## Text Image Compression Using Soft Pattern Matching (1997)

Venue: | Computer Journal |

Citations: | 35 - 9 self |

### BibTeX

@ARTICLE{Howard97textimage,

author = {Paul G. Howard},

title = {Text Image Compression Using Soft Pattern Matching},

journal = {Computer Journal},

year = {1997},

volume = {40},

pages = {146--156}

}

### Years of Citing Articles

### OpenURL

### Abstract

this paper we describe a process which can be used for both lossless and lossy compression. For text documents at 200 dpi, our lossless compression ratios are between 20% and 65% better than those of the JBIG-1 standard [1]. Our lossy compression ratios are between 2.0 and 4.6 times the lossless ratios of JBIG-1, with only barely perceptible changes from the original. The lossless algorithm is similar to the method described by Mohiuddin et al. [2]; we extend the method to allow lossy compression by preprocessing each character in a way that reduces the number of bits output without noticeably distorting the character.

### Citations

831 | Managing Gigabytes - Compressing and Indexing Documents and Images
- Witten, Moffat, et al.
- 1999
(Show Context)
Citation Context ...ive. In the second phase, we find an approximate range in which n lies by testing ranges of size 1, 2, 4, 8, 16, . . . . If n is non-negative, we test whether n is in [0, 0], [1, 2], [3, 6], [7, 14], =-=[15, 30]-=-, . . . , the ranges doubling in size at each step. If n is negative, we test [-1, -1], [-3, -2], [-7, -4], [-15, -8], [-31, -16], . . . . Finally, in the third phase we code the value of n within the... |

677 |
Arithmetic coding for data compression
- Witten, Neal, et al.
- 1987
(Show Context)
Citation Context ... nodes, moving down the tree in ever-increasing ranges. The open circles represent third-phase decisions, traversing a complete binary subtree to reach the specific value of n. bitstream. (See [3] or =-=[4]-=- for a detailed description of arithmetic coding.) It does not matter to the encoder and decoder where the probabilities come from or how they change as long as the decoder has access to the same set ... |

351 |
Universal codeword sets and representations of the integers
- Elias
- 1975
(Show Context)
Citation Context ... is negative. In the second phase, we find an approximate range in which n lies by testing ranges of size 1, 2, 4, 8, 16, . . . . If n is non-negative, we test whether n is in [0, 0], [1, 2], [3, 6], =-=[7, 14]-=-, [15, 30], . . . , the ranges doubling in size at each step. If n is negative, we test [-1, -1], [-3, -2], [-7, -4], [-15, -8], [-31, -16], . . . . Finally, in the third phase we code the value of n ... |

144 | Arithmetic coding revisited
- Moffat, Neal, et al.
- 1998
(Show Context)
Citation Context ...d decoder to maintain a separate set of probabilities for each conditioning class. Binary arithmetic coding Arithmetic coding can deal with alphabets of any reasonable size, up to millions of symbols =-=[5]-=-. However, arithmetic coding tends to be slow except when the alphabet has just two symbols. In the binary case a number of simplifying approximations are possible. The QM-Coder used in the JBIG1 stan... |

113 | Document analysis system - Wong, Wahl - 1982 |

58 | Arithmetic coding for data compression - Howard, Vitter - 1994 |

52 | A Means for Achieving a High Degree of Compaction on Scan-Digitized Printed Text - Ascher, Nagy - 1974 |

23 |
Lossless Binary Image Compression Based on
- Mohiuddin, Rissanen, et al.
- 1984
(Show Context)
Citation Context ...tios are between 2.0 and 4.6 times the lossless ratios of JBIG-1, with only barely perceptible changes from the original. The lossless algorithm is similar to the method described by Mohiuddin et al. =-=[2]-=-; we extend the method to allow lossy compression by preprocessing each character in a way that reduces the number of bits output without noticeably distorting the character. This paper is organized a... |

16 | Combining symbol matching facsimile data compression system - Pratt, Capitant, et al. - 1980 |

11 |
Progressive Bi-level Image Compression
- JBIG
- 1993
(Show Context)
Citation Context ...be a process which can be used for both lossless and lossy compression. For text documents at 200 dpi, our lossless compression ratios are between 20% and 65% better than those of the JBIG-1 standard =-=[1]-=-. Our lossy compression ratios are between 2.0 and 4.6 times the lossless ratios of JBIG-1, with only barely perceptible changes from the original. The lossless algorithm is similar to the method desc... |

10 | A fast binary template matching algorithm for document image data compression - Holt - 1988 |

9 | and C.S.Xydeas, â€śRecent developments in image data compression for digital facsimile - Holt - 1986 |

5 |
at. Textual image compression
- Witten, Bell, et al.
- 1992
(Show Context)
Citation Context ... is negative. In the second phase, we find an approximate range in which n lies by testing ranges of size 1, 2, 4, 8, 16, . . . . If n is non-negative, we test whether n is in [0, 0], [1, 2], [3, 6], =-=[7, 14]-=-, [15, 30], . . . , the ranges doubling in size at each step. If n is negative, we test [-1, -1], [-3, -2], [-7, -4], [-15, -8], [-31, -16], . . . . Finally, in the third phase we code the value of n ... |

2 | Coding of twolevel images by pattern matching and substitution - Johnsen, Segen, et al. - 1983 |