1、Union internationale des tlcommunications Place des Nations 1211 GENEVE 20 Suisse Switzerland Suiza COVERING NOTE GENERAL SECRETARIAT INTERNATIONAL TELECOMMUNICATION UNION Geneva, 16 December 2004 ITU TELECOMMUNICATION STANDARDIZATION SECTOR Subject: Erratum 1 (12/2004) to ITU-T Recommendation T.88
2、(02/2000), Amendment 1 (06/2003), Information technology Lossy/lossless coding of bi-level images, Encoder 1) Clause 10 In the last sentence, replace “7.4“ with “7.4.1“. 2) Clause 11 a) Modify clauses 11 and 11.1 as follows: 11 Page break-up The page break-up (“Front end“) procedures in this clause
3、9 are conceptually the inverse of the page make-up (“Back end“) procedures already described in clause 8. However, pagePage break-up also requires additional page and document decomposition steps prior to encoding. 11.1 Page break-up architectureEncoder model This clause describes the JBIG2 encoding
4、 architectureencoder break-up defined by compliant, but optional, technical components (with a range of algorithms possible to implement each of these components). These JBIG2 page break-upencoding components are a set of processing steps labelled: Capture, Filter, Orient (de-skew), Identify, eXtrac
5、t, Screen, Align (register), Match, Post-match, Dictionary (optimize), and Refine and Encode (bitstream generation). An example sequence of this component set is illustrated in the Encoder Architecturale Components figure below as the horizontal axis with abbreviated labels C F O I X S A M P D R E (
6、leading from input on the left to a compressed data stream on the right). The vertical dimension above each label represents the range of possible algorithms that may be used to implement each component. The horizontal band illustrates an example JBIG2 compliant encoding page break-up method, using
7、some algorithm for each architectural component of this JBIG2 encoding and spanning overall these components. 2 T.88AMD.1_F11.1C O X AM DRPage break-up Encoding procedure Architectural componentsT.88AMD.1_F11.1EncoderNormativeInformativeScopeArchitectureOptionalmethodCFOIXSAMPDREcompresseddataTechno
8、logy componentsA compliant JBIG2 encoder need not include all architectural components, nor use them in exactly the above sequence. b) In 11.2, replace “11.2.12“ with “11.2.11“. c) In 11.2.1, add a bracket as follows: . (e.g., a bitmap rasterized from a document created using a typical word processo
9、r), . d) In 11.2.4, replace “line-arts“ with “line-art“. e) In 11.2.10, replace “any other symbols“ with “more than one mark“. f) Delete clause 11.2.12. 3) Annex J a) Modify the second bullet as follows: minimum resolution of 300 dpi (to reduce matching error and accommodate a number of encoder syst
10、em simplifications, which would otherwise be required for 200-dpi images); b) In the second paragraph, modify last sentence as follows: Selecting one of each method will should result in reasonable encoder performance. 3 c) Modify “J.1 List of JBIG2 encoding components and corresponding algorithms“
11、as follows: Architectural component Norm/ Inform Component method Method reference (Found in J.2) Capture (rasterize) Resolution Striping N Digitize from sensor to bi-tonal bitmap 300 dpi 2 or more None None Filter Fly-speck remove Quantization error remove N Isolated mismatched pixels method Single
12、 protruding pixel method Figure 3 of J1 Page 217 of J1 Orient (de-skew) N Adjacent mark-based slope detection, Hough transform Pages 357-372 of J2 Identify (Region) Segment Classify N Recursive x-y cut Texture analyzer Pages 372-384 of J2 Pages 385-388 of J2 Extract Isolate (black shapes) Truncate (
13、by size) Force (to avoid generic region encoding) N 8 connected boundary trace/rubout-based region-fill approach Min/max of shape w & h Not required (default) Pages 320-325 of J2, J3 J3 None Screen (means of speeding up match) N Comparison of height, width and/or distribution of black pixels Pages 3
14、32-333 of J2 Align (register) N Alignment of symbols using centroid Pages 332-333 of J2 Scanned Match Lossy/Lossless Lossless N Weighted XOR (WXOR) plus CSIS (combined size-independent strategy)b) /CTM (compression-based template matching)c) XOR (Hamming distance = 0) Pages 325-332 of J2, J3 and J4
15、None Post-match Symbol optimize Symbol encoding Symbol placement optimization N Best dictionary symbol shape (not just simple average of the marks) Direct encoding (no refinement) Align symbol bottoms None None None Dictionary Singletons Page spanning N In dictionaryd)(default) Sequentially incremen
16、tale)(default) None None Refine N Arithmetic (inverse of JBIG2 decoder process) J4, ITU-T Rec. T.88 | ISO/IEC 14492 Encode (bitstream generation) N Arithmetic or Huffman (inverse of JBIG2 decoder process) ITU-T Rec. T.88 | ISO/IEC 14492 4 Architectural component Norm/ Inform Component method Method
17、reference (Found in J.2) Capture (rasterize) at page or character level Resolution Striping N Rasterize from generated to bi-tonal bitmap at page or character level 300 dpi 2 or more None None Filter N N/A N/A Orient (de-skew) N N/A N/A Region Identify (Region) Segment Classify N Recursive x-y cut T
18、exture analyzer Pages 372-384 of J2 Pages 385-388 of J2 Extract Isolate (black shapes) Truncate (by size) Force (to avoid generic region encoding) N 8 connected boundary trace/rub-based region-fill approach Min/max of shape w & h Not required (default) Pages 320-325 of J2, J3 J3 None Screen (means o
19、f speeding up match) N Comparison of height, width and/or distribution of black pixels Pages 332-333 of J2 Align (register) N Identical (Bounding box is identical) None Matcha) N XOR (Hamming distance = 0) None Post-match Symbol optimize Symbol encoding Symbol placement optimization N N/A Direct enc
20、oding (no refinement) N/A N/A None N/A Dictionary Singletons Page spanning N In dictionaryd)(default) Sequentially incrementale)(default) None None Refine N N/A N/A Generated Encode (bitstream generation) N Arithmetic or Huffman (inverse of JBIG2 decoder process) ITU-T Rec. T.88 | ISO/IEC 14492 a)Fo
21、r some cases of JBIG2 encoding from Generated data, it may be practical to also eliminate the Extract, Screening and Align Registration components by implementing the Matching component before rasterization, which is typically part of the Capture (rasterize) component. b)Add CSIS to reduce substitut
22、ion errors of WXOR. c)Select WXOR+CSIS for less computation than CTM. d)Minimizing number of Singletons in the dictionary will reduce dictionary memory requirements, and optimizing the sizes (h, w) of Singletons in the dictionary may increase the compression. e)Random page access is best enabled if, for example, one uses a single multi-page-spanning common dictionary, plus multiple page unique dictionaries.
