1DJVU(1) DjVuLibre-3.5 DJVU(1)
2
6 DjVu - DjVu and DjVuLibre.
7
10 Although the Internet has given us a worldwide infrastructure on which
11 to build the universal library, much of the world knowledge, history,
12 and literature is still trapped on paper in the basements of the
13 world's traditional libraries. Many libraries and content owners are in
14 the process of digitizing their collections. While many such efforts
15 involve the painstaking process of converting paper documents to com‐
16 puter-friendly form, such as SGML based formats, the high cost of such
17 conversions limits their extent. Scanning documents, and distributing
18 the resulting images electronically is not only considerably cheaper,
19 but also more faithful to the original document because it preserves
20 its visual aspect.
21 Despite the quickly improving speed of network connections and comput‐
23 ers, the number of scanned document images accessible on the Web today
24 is relatively small. There are several reasons for this.
25 The first reason is the relatively high cost of scanning anything else
27 but unbound sheets in black and white. This problem is slowly going
28 away with the appearance of fast and low-cost color scanners with sheet
29 feeders.
30 The second reason is that long-established image compression standards
32 and file formats have proved inadequate for distributing scanned docu‐
33 ments at high resolution, particularly color documents. Not only are
34 the file sizes and download times impractical, the decoding and render‐
35 ing times are also prohibitive. A typical magazine page scanned in
36 color at 100 dpi in JPEG would typically occupy 100 KB to 200 KB , but
37 the text would be hardly readable: insufficient for screen viewing and
38 totally unacceptable for printing. The same page at 300 dpi would have
39 sufficient quality for viewing and printing, but the file size would be
40 300 KB to 1000 KB at best, which is impractical for remote access.
41 Another major problem is that a fully decoded 300 dpi color images of a
42 letter-size page occupies 24 MB of memory and easily causes disk swap‐
43 ping.
44 The third reason is that digital documents are more than just a collec‐
46 tion of individual page images. Pages in a scanned documents have a
47 natural serial order. Special provision must be made to ensure that
48 flipping pages be instantaneous and effortless so as to maintain a good
49 user experience. Even more important, most existing document formats
50 force users to download the entire document first before displaying a
51 chosen page. However, users often want to jump to individual pages of
52 the document without waiting for the entire document to download.
53 Efficient browsing requires efficient random page access, fast sequen‐
54 tial page flipping, and quick rendering. This can be achieved with a
55 combination of advanced compression, pre-fetching, pre-decoding,
56 caching, and progressive rendering. DjVu decomposes each page into mul‐
57 tiple components (text, backgrounds, images, libraries of common
58 shapes...) that may be shared by several pages and downloaded on
59 demand. All these requirements call for a very sophisticated but par‐
60 simonious control mechanism to handle on-demand downloading, pre-fetch‐
61 ing, decoding, caching, and progressive rendering of the page images.
62 What is being considered here is not just a document image compression
63 technique, but a whole platform for document delivery.
64 DjVu is an image compression technique, a document format, and a soft‐
66 ware platform for delivering documents images over the Internet that
67 fulfills the above requirements.
68
71 The DjVu image compression is based on three technologies:
72 DjVuPhoto
74 DjVuPhoto, also known as IW44, is a wavelet-based continuous-tone image
75 compression technique with progressive decoding/rendering. It is best
76 used for encoding photographic images in colors or in shades of gray.
77 Images are typically half the size as JPEG for the same distortion.
78 DjVuBitonal
80 DjVuBitonal, also known as JB2, is a bitonal image compression that
81 takes advantage of repetitions of nearly identical shapes on the page
82 (such as characters) to efficiently compress text images. It is best
83 used to compress black and white images representing text and simple
84 drawings. A typical 300 dpi page in DjVuBitonal occupies 5 to 25 KB (3
85 to 8 times better than TIFF-G4 or PDF ).
86 DjVuDocument
88 DjVuDocument is a compression technique specifically designed for color
89 digital documents images containing both pictures and text, such as a
90 page of a magazine. DjVuDocument represents images into separately
91 compressed layers. The foreground layer is usually compressed with
92 DjVu Bitonal and contains the text and drawings. The background layer
93 is usually compressed with DjVuPhoto and contains the background tex‐
94 ture and the pictures at lower resolution.
95
98 The DjVu technology is designed from the ground up to support the effi‐
99 cient delivery of digital documents over the Internet. It provides
100 various ways to deal with multi-page documents, and various ways to
101 enrich the content with hyper-links, meta-data, searchable text, etc.
102 MIME types
105 The DjVu format has an official MIME type of image/vnd.djvu, which is
106 the preferred content-type to be given by http servers for DjVu files.
107 Unofficial mime types used historically are image/x.djvu and image/x-
108 djvu, which may still be encountered. Ideally, clients should be con‐
109 figured to handle all three. (For web server configuration help, see
110 http://www.djvuzone.org/support/tutorial/chapter-authoring1.html.)
111 Bundled multi-page documents
114 Bundled multi-page DjVu document uses a single file to represent the
115 entire document. This single file contains all the pages as well as
116 ancillary information (e.g. the page directory, data shared by several
117 pages, thumbnails, etc.). Using a single file format is very conve‐
118 nient for storing documents or for sending email attachments.
119 When you type the URL of a multi-page document, the DjVu browser plugin
121 starts downloading the whole file, but displays the first page as soon
122 as it is available. You can immediately navigate to other pages using
123 the DjVu toolbar. Suppose however that the document is stored on a
124 remote web server. You can easily access the first page and see that
125 this is not the document you wanted. Although you will never display
126 the other pages the browser is transferring data for these pages and is
127 wasting the bandwidth of your server (and the bandwidth of the Internet
128 too). You could also see the summary of the document on the first page
129 and jump to page 100. But page 100 cannot be displayed until data for
130 pages 1 to 99 has been received. You may have to wait for the trans‐
131 mission of unnecessary page data. This second problem (the unnecessary
132 wait) can be solved using the ``byte serving'' options of the HTTP/1.1
133 protocol. This option has to be supported by the web server, the prox‐
134 ies, the caches and the browser. Byte serving however does not solve
135 the first problem (the waste of bandwidth).
136 Indirect multi-page documents
138 Indirect multi-page DjVu documents solve both problems. An indirect
139 multi-page DjVu document is composed of several files. The main file
140 is named the index file. You can browse a document using the URL of
141 the index file, just like you do with a bundled multi-page document.
142 The index file however is very small. It simply contains the document
143 directory and the URLs of secondary files containing the page data.
144 When you browse an indirect multi-page document, the browser only
145 accesses data for the pages you are viewing. This can be done at a
146 reasonable speed because the browser maintains a cache of pages and
147 sometimes pre-fetches a few pages ahead of the current page. This
148 model uses the web serving bandwidth much more effectively. It also
149 eliminates unnecessary delays when jumping ahead to pages located any‐
150 where in a long document.
151 Annotations
153 Every DjVu image optionally includes so-called annotation chunks. The
154 annotation chunk is often used to define hyper-links to other document
155 pages or to arbitrary web pages. Annotation chunks can also be used
156 for other purposes such as setting the initial viewing mode of a page,
157 defining highlighted zones, or storing arbitrary meta-data about the
158 page or the document.
159 Hidden text
161 Every DjVu image optionally includes a hidden text layer that associ‐
162 ated graphical features with the corresponding text. The hidden text
163 layer is usually generated by running an Optical Character Recognition
164 software. This textual information provides for indexing DjVu docu‐
165 ments and copying/pasting text from DjVu page images.
166 Thumbnails
168 DjVu documents sometimes contain pre-computed page thumbnails.
169
172 The DjVu technology was initially created by a few researchers in AT&T
173 Labs between 1995 and 1999. Lizardtech, Inc. (
174 http://www.lizardtech.com ) then obtained a commercial license from
175 AT&T and continued the development. They have now a variety of solu‐
176 tions for producing and distributing documents using the DjVu technol‐
177 ogy.
178 The DjVuZone web site ( http://www.djvuzone.org ) is managed by the few
180 AT&T Labs researchers who created the DjVu technology in the first
181 place. We promote the DjVu technology by providing an independent
182 source of information about DjVu.
183 Understanding how little room there is for a proprietary document for‐
185 mat, Lizardtech released the DjVu Reference Library under the GNU Pub‐
186 lic License in December 2000. This library entirely defines the com‐
187 pression format and the elementary codecs. Six month later, Lizardtech
188 released an updated DjVu Reference Library as well as the source code
189 of the Unix viewer.
190 These two releases form the basis of our initial DjVuLibre software.
192 We modified the build system to comply with the expectations of the
193 open source community. Various bugs and portability issues have been
194 fixed. We also tried to make it simpler to use and install, while pre‐
195 serving the essential structure of the Lizardtech releases.
196 The DjVuLibre software contains the following components:
198 bzz(1) A general purpose compression command line program. Many inter‐
200 nal DjVu data structures are compressed using this technique.
201 c44(1) A DjVuPhoto command line encoder. This state-of-the-art wavelet
203 compressor produces DjVuPhoto images from PPM or JPEG images.
204 cjb2(1)
206 A DjVuBitonal command line encoder. This soft-pattern-matching
207 compressor produces DjVuBitonal images from PBM images. It can
208 encode images without loss, or introduce small changes in order
209 to improve the compression ratio. The lossless encoding mode is
210 competitive with that of the Lizardtech commercial encoders.
211 cpaldjvu(1)
213 A DjVuDocument command line encoder for images with few colors.
214 This encoder is well suited to compressing images with a small
215 number of distinct colors (e.g. screen-shots). The dominant
216 color is encoded by the background layer. The other colors are
217 encoded by the foreground layer.
218 csepdjvu(1)
220 A DjVuDocument command line encoder for separated images. This
221 encoder takes a file containing pre-segmented foreground and
222 background images and produces a DjVuDocument image.
223 ddjvu(1)
225 A command line decoder for DjVu images. This program produces a
226 PNM image representing any segment of any page of a DjVu docu‐
227 ment at any resolution.
228 djview(1)
230 A stand-alone viewer for DjVu images. This sophisticated viewer
231 displays DjVu documents. It implements document navigation as
232 well as fast zooming and panning.
233 nsdejavu(1)
235 A web browser plugin for viewing DjVu images. This small plugin
236 allows for viewing DjVu documents from web browsers. It inter‐
237 nally uses djview to perform the actual work.
238 djvups(1)
240 A command line tool for converting DjVu documents into Post‐
241 Script .
242 djvm(1)
244 A command line tool for manipulating bundled multi-page DjVu
245 documents. This program is often used to collect individual
246 pages and produce a bundled document.
247 djvmcvt(1)
249 A command line tool for converting bundled documents to indirect
250 documents and conversely.
251 djvused(1)
253 A powerful command line tool for manipulating multi-page docu‐
254 ments, creating or editing annotation chunks, creating or edit‐
255 ing hidden text layers, pre-computing thumbnail images, and
256 more...
257 djvutxt(1)
259 A command line tool to extract the hidden text from DjVu docu‐
260 ments.
261 djvudump(1)
263 A command line tool for inspecting DjVu files and displaying
264 their internal structure.
265 djvuextract(1)
267 A command line tool for dis-assembling DjVu image files.
268 djvumake(1)
270 A command line tool for assembling DjVu image files.
271 djvuserve(1)
273 A CGI program for generating indirect multi-page DjVu documents
274 on the fly.
275 djvutoxml(1), djvuxmlparser(1)
277 Command line tools to edit DjVu metadata as XML files.
278
281 DjVuLibre comes with a variety of specialized encoders, c44(1) for pho‐
282 tographic images, cjb2(1) for bitonal images, and cpaldjvu(1) for
283 images with few distinct colors. Although these encoders perform well
284 in their specialized domain, they cannot handle complex tasks involving
285 segmentation and multipage encoding.
286 The Lizardtech commercial products (see http://www.lizardtech.com/solu‐
288 tions/document) can perform these complex encoding tasks
289 Another solution is provided by the compression server at
292 (http://any2djvu.djvuzone.org). This machine uses pre-lizardtech pro‐
293 totype encoders from AT&T Labs and performs almost as well as the com‐
294 mercial Lizardtech encoders. Please note that the Any2DjVu compression
295 server comes with no guarantee, that nothing is done to ensure that
296 your documents will remain confidential, and that there is only one
297 computer working for the whole planet.
298
301 Numerous people have contributed to the DjVu source code during the
302 last five years. Please submit a sourceforge bug report to update the
303 following list.
304 Yoshua Bengio, Léon Bottou, Chakradhar Chandaluri, Regis M. Chaplin,
306 Ming Chen, Parag Deshmukh, Royce Edwards, Andrew Erofeev, Praveen
307 Guduru, Patrick Haffner, Paul G. Howard, Orlando Keise, Yann Le Cun,
308 Artem Mikheev, Florin Nicsa, Joseph M. Orost, Steven Pigeon, Bill
309 Riemers, Patrice Simard, Jeffery Triggs, Luc Vincent, Pascal Vin‐
310 cent.
311DjVuLibre-3.5 10/11/2001 DJVU(1)