Scan and Share 1.

07-st
Tutorial on making e-books
written by V. and A.
2010

Contents
1 Introduction

1.1 In brief . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2 Why make a scanned book, is OCR not good? . . . . . . . . . . . .

1.3 How to get good quality of scans . . . . . . . . . . . . . . . . . . . .

2 Scanning a book

2.1 Setting up IrfanView for scanning . . . . . . . . . . . . . . . . . . .

2.2 Setting up VueScan for scanning . . . . . . . . . . . . . . . . . . .

2.3 Handwork while scanning . . . . . . . . . . . . . . . . . . . . . . .

3 Processing scans with ScanKromsator

13

3.1 Draft run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.2 Set options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3 Main run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.4 Processing color figures and photos . . . . . . . . . . . . . . . . . . 18
4 Processing scans with ScanTailor

19

4.1 Importing scan into ScanTailor . . . . . . . . . . . . . . . . . . . . 20

4.2 Draft run . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.3 More about processing steps . . . . . . . . . . . . . . . . . . . . . . 22
4.4 Correct errors after the draft run . . . . . . . . . . . . . . . . . . . 23
4.4.1 Adjusting the content rectangle . . . . . . . . . . . . . . . . 25
4.4.2 Adjusting the page alignment

. . . . . . . . . . . . . . . . . 25

4.4.3 Adjusting the page sizes . . . . . . . . . . . . . . . . . . . . . 26

1

4.4.4 Adjusting the splitting . . . . . . . . . . . . . . . . . . . . . . 26

4.4.5 Adjusting the deskewing . . . . . . . . . . . . . . . . . . . . 28
4.4.6 Replacing scans in the project . . . . . . . . . . . . . . . . . 28
4.5 Final run and final check-up . . . . . . . . . . . . . . . . . . . . . . 29
4.6 Working with picture zones . . . . . . . . . . . . . . . . . . . . . . . 31
5 Encoding scans into DJVU

33

6 Creating text layer with OCR

36

7 Adding book covers and color plates

38

8 Adding hyperlinks and bookmarks

38

A Where to download software

40

This document can be distributed for free. It is an expanded version of the

Scan and Share 1.07 tutorial. This tutorial now covers the new program,
ScanTailor, as well as ScanKromsator. Some screenshots are in Russian because the software does not have any other localization. Screenshots of VueScan options are included now.

1 Introduction
This is a mini-tutorial about scanning books and making high-quality files
out of them. This tutorial is intended for people who would like to make goodquality electronic books but do not know where to start. There are many
ways to get good results by scanning; this text shows you some reasonably
easy ways. The tutorial has step-by-step screenshots and assumes some familiarity with Windows. You may need to download and install a few programs
(see Appendix A).

1.1 In brief
For the impatient reader: The process consists roughly of the following stages:
1. Scan every page in 300dpi greyscale, save to TIF. Save a backup of your
scans!
2. Import images into ScanKromsator or ScanTailor, process images. Save
a backup of the processed images at this stage!
3. Create a DJVU file out of processed images.
4. Add OCR and/or bookmarks to the DJVU file.
(It is most important to master the stages 1 and 2, since the processed images
after stage 2 are much smaller than the initial scans, and you can send them
to somebody else if you have trouble with stages 3 and 4.)

1.2 Why make a scanned book, is OCR not good?

Here I will be mostly talking about scanning of old books on science, mathematics, or technical books. For these books, OCR is not practical because
these books contain too many equations, diagrams, graphs etc. No OCR program can accurately recognize this kind of material. The only solution is to
scan and make images of all pages.

1.3 How to get good quality of scans

Such books are almost always printed purely in black/white, with perhaps
very few pages having greyscale or color illustrations. For that kind of books,
the highest quality of scanned e-books is achieved if one uses 600dpi black/white
images for most pages.1 So you need to scan either directly in 600dpi black/white,
or at 300dpi greyscale and then process the scans to make them into 600dpi
1

If you dont know what 600dpi means: its called the resolution of the image and means
the number of image points (pixels) per inch (dpi=dots per inch).

black/white.2 If the book has a few pages with color illustrations, you will
need to scan them separately in 300dpi 24-bit color mode. The same applies
to colorful book covers that you also may want to scan.
Please note:
Never scan at 300dpi black/white! The quality of the results is never as
good as what you can get by scanning in 300dpi greyscale and following
this tutorial or equivalent methods.
Scanning in 300dpi greyscale is on most scanners exactly as quick as
scanning in 300dpi black/white or in any lower resolution! You will
not save time if you scan in 300dpi black/white or in 200dpi instead of
300dpi greyscale, but you do lose a lot of quality.
Scanning in 300dpi greyscale produces large intermediate scanned files,
which will be processed into very small DJVU files. Scanning in 600dpi
black/white produces smaller intermediate scanned files, but the process of scanning at 600dpi is much slower for most scanners. Also, its
easier to process 300dpi greyscale scans because they have less "digital
dirt" than 600dpi black/white scans.
It is nearly impossible to improve the quality of a poorly scanned and/or
incorrectly processed image of a book. For example, some e-books are
made by inexperienced people in 150dpi, or in color instead of black/white,
or the resolution was decreased after scanning in an attempt to reduce
the file size. These e-book files are huge in size. The visual and print
quality of such e-books is bad and cannot be improved! It is important
(and not difficult) to make the scanned image correctly and ensure great
quality of the resulting e-books. Read on!
A high-quality scanned e-book is small in size, has great visual appearance
on the screen and also when printed, and has searchable text. There are
many ways to achieve high quality of scanned e-books; all methods involve the
resolution of 600dpi. (Higher resolution almost never brings a significantly
better quality.) Output files are in the DJVU3 format and take typically about
5KB/page to 10KB/page. If your file is significantly larger, while the book
contains only black/white text and is printed reasonably clearly, something
was done incorrectly when producing the file.
You may of course experiment on your own with other programs. For example, some people use Photoshop with special plugins, Book Restorer, Corel
PhotoPaint, RasterID, even Matlab and IDL for image processing. This tutorial
presents a particular method that practically guarantees good results. If you
are a beginner, please make a few books by closely following the instructions
2

This kind of processing when the resolution of an image is increased is called upsampling.
If you dont know what DJVU is, please use Google or Wikipedia to read about it. The
DJVU format was specially developed for high-compression storage of scanned images. Most
e-books today are in the PDF format, but the PDF format was intended for documents created
in a word processor, i.e. for vector documents rather than scanned documents. Scanned ebooks in PDF format occupy more space and/or display slower than in the DJVU format.
3

in this tutorial. You will then see that you can achieve quite a high a level of
quality without excessive effort and without learning too many technicalities.
If you develop your own methods, for example by using different options or
different programs, you will be able to decide which method is best because
you can then compare the quality of the results with the reference quality
obtained by the methods in this tutorial.

2 Scanning a book
You pick up a thick volume. Maybe you think that only a maniac could scan
it, page after page. Yes, you are right! But you can become that kind of
maniac and scan books of any size without much discomfort if you organize
your work well.
For the impatient reader:
use any flatbed scanner, even a cheap one, and a program such as
IrfanView to control scanning
do not use a digital camera for scanning books!
do not use FineReader for scanning books!
Why not use FineReader for scanning? The FineReader is a good program
for making OCR but is not optimal for scanning and for processing the scans
with the goal of making a scanned e-book. FineReader attempts to give you
a kind of all-in-one solution for scanning and processing e-books; please resist the temptation to use just one program for everything. You will not get
good results with FineReader; in any case, nowhere as good as when you
follow this tutorial. FineReader has the following technical drawbacks: 1)
It sometimes uses JPEG for image compression. This is not appropriate for
black/white text! 2) It stores images internally as black/white 300dpi TIFFs
and auto-rotates them. Black/white 300dpi is adequate for OCR but not optimal for digital scanned e-books. The auto-rotate algorithm is faulty and
produces defects in the image (broken lines). The auto-rotation is hardcoded into FineReader 7.x, 8.x and cannot be disabled.4 3) If you scan in
300dpi greyscale, which is the procedure recommended here, FineReader will
perform all operations at 300dpi, rather than resample to 600dpi. ScanKromsator and ScanTailor will first resample to 600dpi and then perform processing. The results of FineReader processing are always going to be inferior for
these reasons.
Why not use a digital camera for scanning books? You will never get good
results even with expensive 10 Megapixel or whatever cameras. Never even
closely as good as with a flatbed scanner, even a cheap one. Look at figure 1
below and guess which of the two images of the same page is made by a digital
camera.
4

Only in FineReader version 9 there was added an option to disable this auto-rotation.
However, other features of FineReader remain. Also not ethat FineReader version 9 cannot
be used to produce OCR layer in DJVU files. I recommend using FineReader version 8.

Figure 1: Two images of the same page, one made by a digital camera, another by a cheap flatbed scanner. The image made by a flatbed scanner was
scanned at 300dpi greyscale and upsampled to 600dpi black/white. You can
guess which image was made by the digital camera! (Yes, the crappy one.)
We recommend that you always use a flatbed scanner and scan at 300dpi
greyscale or higher resolution.

For scanning, you need basically any program that can work with your scanner. Under Windows, the TWAIN scanner driver is popular.5 Under Linux,
many scanners are supported by the VueScan program, but you can use any
other program as long as your scanner is supported.
You can scan using any program like IrfanView, XnView, ACDSee, PhotoShop.
(Note that IrfanView is small an free.) It is important that your scanning
program does not try to do anything with the images; in particular, no deskew,
no optimizing, no resizing, nothing at all. You should be able to tell the
program just to save the scans for each page to the hard disk in the TIF
format.
It is convenient if your scanning program can save scanned images for every
page one after another, numbering the files like p0001.tif, p0002.tif, etc. For
example, VueScan and IrfanView can do this.

2.1 Setting up IrfanView for scanning

As an example, we describe how to scan using IrfanView. (This program can
be downloaded for free.) Scanning in other programs is quite similar.
Start IrfanView. In the File menu, press "Choose TWAIN Source". Choose the
scanner that you need to use.

Then in the same menu choose "Acquire/Batch scan".

Most scanners are supported by TWAIN drivers; for other scanners you may need special
drivers.

Here you can choose how to number the scanned files, where to store them,
and in which format to save them. As shown, the files will be named page0001.tif,
page0002.tif, etc. You should select TIFF as the image format. (Do not use
JPEG as the output format!)
Click on Options to the right of Save as field. This will set the options for the
TIFF format.

You should select LZW compression; this will cut the TIFF file size in two,
compared with no compression (None).6 If you later find that you have compatibility problems with these TIFF files (i.e. you later use a program that
cannot open them) then you need to change the compression method.
Important: Do not use the JPEG compression method for black/white text!
JPEG compression introduces digital artifacts, that is funny-looking shades
around each letter (see figure 2). It is pointless to use JPEG for black/white
images.7
Now press OK and go to the TWAIN driver window for your scanner.
In the TWAIN window (or other configuration window if you are not using
TWAIN drivers), set the resolution to 300dpi and the color mode to greyscale.
(In some programs, this is called 8 bit greyscale.) These are the most important settings. Some scanning programs do not allow you to set explicitly the
resolution or the color mode; instead they say something like Black/white
photo or web-optimized quality. Avoid these programs, instead use some
program that allows you to set specifically 300dpi and 8-bit greyscale. If you
are not sure that your settings are right, you should try scanning one page,
save the file to disk as TIF, and check the properties of the file in a graphics
editor, to make sure that you actually got 300dpi and 8-bit greyscale.
6

Note that a typical page scanned in greyscale will occupy between 2 and 4 megabytes on
the hard disk with LZW compression.
7
The JPEG format actually cannot handle black/white images; when one converts
black/white images to JPEG, the software must convert those images into greyscale images.
The JPEG compression then introduces a certain quality loss, as shown in the figure. The
quality loss in JPEG compression is acceptable for photographs but may degrade black/white
text quite significantly, unless a high quality JPEG mode is selected. (The quality of JPEG
compression is usually selectable as a number from 1% to 100%. No visible artifacts would
appear at 90% quality or higher. But some programs, especially for making PDF files or for
optimizing images, may not allow you to set the JPEG quality manually.)

Figure 2: Digital artifacts appearing due to JPEG compression of black/white

text. (In this example, the quality setting for the JPEG encoding was very
low, so these artifacts are apparent to the eye.) At left: greyscale image with
unnatural wavy-looking shadows around the letters. These digital shadows
are typical for JPEG compression of black/white images. At right: the same
image converted back to black/white. The digital artifacts produce digital
noise i.e. speckles around the letters and a distortion of the shape of the
letters.

2.2 Setting up VueScan for scanning

VueScan runs under Linux, Windows, and Macintosh, is not a free program,
but all upgrades are free once you buy it. An advantage of VueScan under
Linux is that it supports many types of scanners that are otherwise not supported by standard Linux software.
In VueScan there are many tabs with options. The first tab (figure 3 left) is
the Input tab that controls the scanning mode. Note that VueScan may not
show you all these options unless you enable the Expert mode (or show all
options). You can make the settings as shown; for instance, you explicitly set
the resolution to 300dpi and the color mode to 8 bit greyscale. It is important
to check the box Lock image color so that each page is scanned in the same
color balance. If you want, you can make automatic scanning with a small
delay; then you will have to jump to the scanner every time to change the
page. I prefer not to do this.
The second tab (figure 3 left) is the Output tab. There you can set the directory where the scans will be kept, the format of file names (in this example, it
will be p001.tif, p002.tif, etc.), and the TIFF compression (On).

2.3 Handwork while scanning

By now you have set up your scanning program. The actual work while scanning is not complicated:
First you need to try scanning some place in the book and check that
everything works well. Take a book, open somewhere where the pages
are full of text, put the book (both pages down) on the scanner glass.
If necessary, press with your hand so that the crease of the book is as
close to the glass as possible. (You can also use a weight, e.g. put another
heavy book on top, but its slower than pressing by hand.)
9

Figure 3: Options for scanning when using VueScan.

10

 Do a preview scan. Then you can see what has been scanned in the
preview window. If needed, you can turn the page 90 degrees so that
the text is straight up. You can also adjust contrast, brightness, gamma
correction if necessary. Your goal is that the text must be clearly visible,
not too dark, and not too light.
Select the scanning region by using the mouse. You should select the
scanning region such that some white space is left around the text, but
no book crease or off-page regions are scanned. Your purpose is that the
scanning rectangle should fit around the text with some margin, so that
you will not lose any text even if you put the book a little askew on the
glass. And yet you do not want to scan any useless regions outside of
the page.
Press the Scan button with the mouse and wait until the scanner finishes scanning the page. This will get the scan of one page (or two pages
at once, if you can fit the book onto the scanner). The scanned file will
be saved to the disk.
Now that the scanning program is set up, you can scan all the pages
with the same settings. While the scanner lamp is moving back, turn
the next page and put the book back to the same place on the scanner.
Then press the mouse button to scan again. (The mouse can be left
pointing at the Scan button, so you dont need to look. Alternatively,
some scanners have buttons on them that make the next scan.)
This technique allows you to scan the entire book, one page after another,
without looking at the computer screen or at the keyboard. You can watch TV
or whatever while you are scanning. Depending on the scanner speed, you
can get between 100 and 200 scanned pages per hour. Some scanners are
particularly fast (e.g. Plustek OpticBook).
It is not necessary to set the book onto the scaner absolutely straight (edge
of the book parallel to the edge of the scanner). You should try to put it
reasonably straight, but it is unavoidable that pages will not all be scanned
completely straight; many pages will be slightly skewed. This small skew is
okay and will be corrected later (after scanning) by software. Correcting this
skew is called deskewing. Deskewing is very fast and efficient.
What you want to avoid when scanning:
Avoid very large skew angles, i.e. do not place a book at a large angle on
the glass. This kind of scan can still be deskewed, but the shapes will
probably not be as smooth as otherwise.
Avoid incomplete page scans, i.e. when some of the text is outside of the
scanning region. This means that some text will be lost (not scanned
at all). If you discover such a page, scan that page again with a correct
scanning region. In a science book, no part of the text is unimportant!

11

However, avoid scanning the library stamps or other marks on the pages. If
your book has stamps or other markings on some pages, just cover them with
a piece of paper while scanning, or remove them with digital image editor after
scanning. Nobody wants to see some ugly stamps or marks in the e-book!
Avoid scanning any off-page regions (this will be when your scanning
rectangle is way too large). This will produce a black shadow which,
in many cases, you will have to remove by hand while processing your
scans! (This is so because computers are not very good at guessing what
is a part of the book and what is dirt on a scan.)
Also, avoid producing a fuzzy image because some place on the page was
not close to the scanner glass.
The region of the text around the book crease is often difficult to scan. You can
try scanning one page at a time (rather than two pages) or pressing slightly
harder onto the book binding. It is important that the text is directly next
to the scanner glass. Even 1 mm distance between the glass and the paper
will make a very fuzzy scanned image in almost all scanners! Fuzzy scanned
images are not acceptable. It is very difficult to prepare a good quality final
e-book from fuzzy scans.
Should you scan one page at a time, or two pages at a time? It is faster to
scan a book two pages per scan rather than one page at a time. Doublepage scans can be cut quite efficiently and automatically (if they are scanned
cleanly) by software. But not all books can be scanned that way; many books
are too large (you wont fit two pages onto the glass unless you have an A3
scanner, which is usually expensive). Many books dont open sufficiently to
be scanned two pages per scan with good quality (some text near the crease is
lost or becomes too fuzzy, which is not acceptable!). You need to try two pages
at a time, try one page at a time, and then decide how to proceed. Regardless
of how you scan, the processing software will be able to prepare an e-book
with single page images, as long as everything is scanned correctly and the
images are clear.
The result, after scanning the entire book, is a directory full of TIFF files.
These files are the raw material that you will start processing after you finish scanning. Note that you need sufficient disk space to store all those
scans (at least 4MB per scanned image!). After you finish scanning, use a
slideshow mode of some picture viewer to quickly preview the scanned images
to make sure that you didnt miss any pages and that every page is adequately
scanned. It will be too late when you discover, at the final processing stage,
that some pages are only half-scanned or missing, especially when the book
has already left your hands!
Note: When you scan the book, scan all pages; please do not omit any pages,
including title pages, front matter, including any information about the publisher, the table of contents, the index, the bibliography, empty pages in the
middle of the book, page numbers, errata sheets, or anything else!!! You will

12

not save much time if you decide to skip 20 pages or so while scanning. However, a science book is almost unusable without bibliography and index and
without exact information about its publication. 8

3 Processing scans with ScanKromsator

Now we discuss the processing software. The first program is the wonderful
ScanKromsator written by Bolega.9 ScanKromsator is a very powerful tool
for processing scanned material. ScanKromsator has a very large number of
useful functions, but some of them are not intuitive or difficult to understand
if you just look at the user interface.10 In this tutorial you will be walked
through a particular simplified workflow with ScanKromsator, assuming that
you scanned a book at 300dpi greyscale.
Start ScanKromsator and load the raw TIFF files into it (menu File). The list of
files will appear on the top left column. The toolbar with several tabs (Book,
etc.) will appear below the list of files.

In the example shown, a book was scanned with two pages per scan, and
apparently there was some skewing. Our task now is to split, to deskew, and
8

Also, do not think that you will make your life easier from the legal point of view if you
dont scan the publication information!
9
Please do not write email to Bolega asking for help, for documentation, for source code
of ScanKromsator, or for adding extra features! Instead, just learn to use it and make some
good quality e-books!
10
We will talk only about the bare minimum of ScanKromsator functions here. Unfortunately the ScanKromsator program does not yet have a comprehensive users manual describing all the functions.

13

to cut the page images so that every page has the same size and margins. If
your scan is single-page, you will not need to split, but you will still need to
deskew and cut. This operation is called kromsating in the program.11

3.1 Draft run

The first step is a draft processing run, i.e. preparation for the final processing
of the raw files.
Click the tab Files in the toolbar. You get a dialog where
you can set the output resolution (very important!) to 600dpi,
the folder for storing the output files (the output folder is by
default the subdirectory out in the current directory), and
the way of numbering the output files (prefix, number of digits, starting number, step). Note the format for compressing
the output files: its TIFF G4 encoding, which is optimal for
black/white TIFF images. This will be the output format after
processing.
To start the draft processing run, click the
button Draft kromsate bearing the pictogram of scissors, which is located to the
left of the Process button in the toolbar.
When you press the Draft kromsate button, and you get the dialog shown at right.
In this dialog you need to set tick marks on
Split pages and Safe top/bottom. The
field Kromsate=All means that the options are applied to all the pages. If some
pages do not need to be split, you can select Kromsate=Current and unset Split
pages for these pages.
Press OK and wait 10-15 minutes until the Draft kromsate operation is
finished. You will get the following screen.
11

The pseudoword kromsate is a mangled Russian word meaning to cut in pieces.

Within the ScanKromsator, the meaning of kromsate is the operation of splitting a two-page
scanned image into individual page images, and also the operation of cutting page images so
that the margins become even and equal on all pages.

14

Note that there are now green tick marks in the page list (top left column),
meaning that these pages have been draft kromsated successfully. For each
page you will see the blue lines across the page. These lines are the cutters that determine how the page image will be cut and split. Note that the
program attempts to determine automatically where to cut the margins and
where to split a two-page image into single pages. In some cases the program
may make a mistake and cut too much or too little; in that case you will later
be able to adjust the position of the cutters by hand.

3.2 Set options

The next important step is to go through the processing options and prepare
for the main (not draft) run of ScanKromsator. The processing options are
set in the many different tabs in the toolbar (left middle column).
Please note: Each option can be set either to apply to all pages at once, or only
to the currently shown page. To apply an option to all pages, hold the Ctrl key
while clicking the option box with the mouse. In this way, you can set some
common options quickly for the entire task and then go to some problematic
page and select other options just for that page.

15

First click the Page tab. Here you can set processing options
for cutting the pages. The option Split means to split the
two-page image into single pages. Deskew will deskew each
single page image separately. Despeckle removes small dots.
Sometimes Deskew makes pages significantly skewed; this
is usually due to some complicated illustrations. In that case,
check Art for these pages. You can set Ortho if the page
needs to be rotated by 90 degrees. You can set these options
separately for left and right (L and R) pages.
Now click on the Book tab. Here you set options related to
the size and layout of the pages in the final book. H.Gap is
the size of the margins. The value of 200 is good for 600dpi
(meaning 1/3 inch). Page width and height can be set to Auto.
You can also center the pages differently (align to center/align
to top/align to bottom).

We already visited the Files tab at the draft stage. It is very important to
have 600dpi as the output resolution in the Files tab!
Now click on the Options tab. Set Deskew method =
Auto (shear), Resample filter = Lanczos3. The setting Despeckle=Fine+Normal or Safe switches on an intelligent despeckle method that avoids removing the dots over i or j,
for example. Text sensitivity controls the logic of the autocutting. Low sensitivity might cut off the page numbers if they
are too far away from the text. You may need to adjust the
sensitivity settings a little bit; but in most cases they do not
need to be adjusted.
You can skip the Options 2 tab for now. Click on the Convert tab. Here you set the threshold for converting greyscale
images to black/white. Do not forget to hold the Ctrl key (to
set this for all pages) as you select Threshold=MiddleDark.
Experiment with other settings if you dont like the results.
Click the Quality tab; there you can further control the conversion to black/white. This is a very important function! Set
Enhance image, Blur=1, and Sharpen=1. What is important
is that the image will become smoother with this setting. The
values of Blur and Sharpen could be 2 instead of 1, although
the value 1 is usually good. A larger value will make the letters more black. You may need to experiment depending on
the quality of printing in a particular book.
Another important option is Gray enhance. Click on it since
you have greyscale scans (which is what you should have!).
16

You will get a dialog with many options for

greyscale images. Go to the Background
cleaner tab and check Enable.
Skip several tabs and click the Illumination
tab; click Correct illumination. This will normalize the illumination of the page, which is
important since usually some parts of the page
are darker than others. This is a very useful feature that removes black shadows that
would otherwise appear in darker places on
the page!

Skip several tabs and click the Denoise tab. Set the parameters as
shown at right. These parameters
clean up the image. This is the last
set of options that we are going to
bother with right now.
You can use the FileOptions... menu to write the options to a file. This will
save you all this work for the next time.
The last step before the main processing is a visual checking of the position
of the cutters. You need to go through every page and check that the cutters
are correctly positioned. Yes, this is a bit boring... but you can make it quick.
Put two fingers of the left hand onto the keys q and w; pressing these keys
will go to the previous/next page. With the right hand, you hold the mouse
and adjust the position of the cutters wherever needed. Sometimes there is a
skewed shadow, or it is necessary for some reason to set the cutter line at an
angle rather than vertically or horizontally. Hold the Shift key and drag the
cutter by its end to achieve this.
You can copy the cutter position from
one page to another. Right-click on the
cutter, and you will see the menu as
shown. For instance, if the current
cutter position needs to be applied to
all subsequent pages, click Copy current position toall down.
17

If some page contains a photograph or a color figure, you need to protect it

from converting to black/white. This can be done when checking the position
of the cutters. Basically, you can select some arbitrary part of the page and
mark it as a picture zone. See Section 3.4 for more details.
You can save the settings for this task by using the File/Save Task command
in the menu. This command is useful if you want to stop the task and to
continue it later.

3.3 Main run

Now that everything is ready, you can begin the main run of ScanKromsator.
Press the large button that says Process and bears the icon of a book, in the
main toolbar at top:

The program will ask you to confirm that you really are sure you want to
change the resolution of the images. Confirm! The process will then start.
Now you need to wait a while. The upsampling operation can be quite slow;
in recent versions of ScanKromsator (5.8 and up) this operation was made
faster. You may expect to process 5 pages per minute or so. When everything
is finished, you should view the output files in the output folder. You should
check that all pages are cut and deskewed correctly. If some pages are not
processed correctly, you can repeat processing of just those pages with some
other options.
The main processing run may take some hours on a slow computer. It is not
necessary to process the entire book in one run. One can process only some
portion of the pages; then one needs to set BookPage widthFixed to the
size determined in the previous portion of the pages (so that all pages have
equal size at the end of processing). It is usually sufficient to take 10 to 15
pages for determining page size.
If you like, you can use the powerful cleaning features of ScanKromsator to
remove the digital dirt from some pages. Typically, the digital dirt is any
extraneous spots on the paper, pencil or pen marks, and library stamps. Of
course, you can also use any graphics editor to clean the images by hand.
Hopefully, there will not be many pages to clean.

3.4 Processing color figures and photos

We discuss color figures separately because they are not frequently needed.
However, their place in the workflow is at the point where you check and
adjust the position of the cutters.
The latest version of Kromsator (5.9) includes a feature for color figure processing, the so-called picture zones. One some pages there may be a picture,
18

i.e. a non-black-white illustration such as a photograph or a colorful diagram.

You need to protect these illustrations from converting into black/white. To
mark a picture zone, select a rectangle containing the illustration and click
on the button Mark as picture zone bearing the icon of a blue frame in this
toolbar:
There is also a possibility to have polygon-shaped picture zones. This is useful, for example, if the page was scanned with a large skewing. Use the starshaped tool button to mark such zones:
To set the options for a picture zone, double-click on the selected region. You
will see the dialog Picture zone properties.

You need to set the color of the illustration. For example, if the page contains
a greyscale photograph (rather than a color photograph or color diagram), set
Color=Gray.
We cannot discuss other zone options here; as you see, there are many options
intended for advanced users. But note that after kromsating the picture
zones will be saved to separate files. So after the main processing run you
will have to merge them with the page files. This is done by using the menu
command: ZonesPicture zoneMerge zones. The resulting page files will be
TIFF files in which the text is black/white but the picture zones have color.

4 Processing scans with ScanTailor

ScanTailor is a relatively new program that is being actively developed; I describe version 0.9.8 at this time. It can be downloaded for free and runs under
Windows and Linux.
The functionality of ScanTailor is sufficient for processing books that have
black/white text and some greyscale illustrations, as well as occasional color
pages. ScanTailor can deskew and clean up your scanned pages, split doublepage scans into single pages, and convert from 300dpi greyscale into 600dpi
black/white, while keeping greyscale illustrations.
19

ScanTailor has online documentation at its website; you can read about many
features of ScanTailor there. Therefore, here I will only show how to do the
most common processing steps.

4.1 Importing scan into ScanTailor

ScanTailor takes as input a number of TIFF files, and produces as output a
new set of TIFF files. When you run ScanTailor, it first asks you to start a new
project or to open a previous project. A project means a bunch of TIFF
files that are going to be processed together. So you say new project at this
point (figure 4).

Figure 4: ScanTailor asks to create a new project or to open a previously

existing project.
Then you will see a dialog box asking you to select the input files (figure 5).
Press Browse on Input directory and select the directory where you have
your scanned TIFF files. (The output directory will be automatically selected
as the out subdirectory. For example, if your scans are in C:\myscans\ then
the output TIFF files will be in C:\myscans\out.) You can now use the arrow
buttons (<< and >>) to exclude some of the TIFF files from processing. You
probably want Select all at this point (i.e. use all the TIFFs in that directory).
Then press OK; the TIFF files will be inspected and a project will be created.
If some of the TIFFs do not have the correct resolution stamped inside them,
you can correct it (fix DPIs), but normally this is not necessary.
After this, you see the main window of ScanTailor that looks like the following:
The selected page is shown in the central window; thumbnails of all pages
are shown in the column at right; and the processing sequence (which I will
explain shortly) is shown on the left.
ScanTailors projects can be saved to files with the extension .scantailor;
these files are in the XML format and have the full information needed to
process the input TIFF files from the project and to produce the output files.
So it is advisable to save the project, also while working in ScanTailor. So you
go to the File menu and choose Save and specify the location and the name
of the project file; for example myscan.
It is also advisable to make the ScanTailor window maximized to full screen;
but I will keep this window small in my examples, just to make screenshots
smaller in this PDF file.

20

Figure 5: ScanTailor asks you to select the input files.

4.2 Draft run

Now that your scans are loaded into ScanTailor, you can start processing.
The optimal way of processing is to let ScanTailor run automatically for all
pages and then correct the errors that may have been made. Even when the
scanned material is very simple and no user interaction is really needed, it
is necessary to have a draft run and a final run because the final output
cannot be produced until all final page sizes are known, and the page sizes
are computed only after the page layout step is performed on all pages.
For the draft run, I suggest the following procedure that seems to be quickest:
You already have the first page selected when you open the project for
the first time. Press with the mouse on the Page layout step (or simply
press P on the keyboard) and wait a little. The first page will be processed through all the steps 1-5 and then you will see the page layout dialog (figure 7). You will see that the first page has been really processed:
deskewed, split, and a content rectangle was selected (everything outside
the content rectangle has been cut away). Dont worry about any options
at this point.
If your scans contain double pages, automatic splitting is what you need.
But if your scans contain only single pages and never need to be split, it
is perhaps better to disable splitting. Click on Split Pages, then click
on Change..., then select Mode: Manual and Scope: All pages. This
will effectively disable automatic splitting for all pages. Now click again
on 5 Page Layout.
Now press the play button to the right of 5 Page Layout.
21

Figure 6: ScanTailors main window with some scans loaded into a project
Unnamed.
This will start the automatic (batch) processing of steps 1-5 for all pages with
the default options. This process will take maybe 20 minutes or so (maybe
about 5 seconds per page), but at least you dont have to do anything while
the program is working. This is your draft run. While it is running, let me
try to explain what is actually happening now.

4.3 More about processing steps

The idea of ScanTailor is to divide the processing into steps as shown in the
left of figure 6. Each step requires that all the previous steps are already
performed on a given page. There is (in version 0.9.8) no way to omit some
steps entirely from processing. You will have control over each step of the
processing and can in principle adjust the settings for each page separately
or apply special settings to a group of pages.
The first step is fix orientation. Here you can rotate pages by 90 degrees or
by 180 degrees, so that the text on the pages is more or less upright. This
step is completely manual; the user needs to supply the rotation for each
page or for all pages. (In order to apply some option for all pages, usually you
need to press the Apply button and then select Apply to all pages.) You
will not need to control this step at all if you adjusted the page orientation
correctly while scanning. By default, ScanTailor will not do anything at this
step. However, you may go to a single particular page (choose it by clicking
on the thumbnail in the right column) and change the orientation if needed.
The second step is split pages. In the example shown in figure 6 there are
22

double-page scans that are already correctly oriented. Most likely, ScanTailor
will automatically and correctly split them into single-page scans. In some
rare cases the splitting is done incorrectly (e.g. too much text is cut off). In
this case you can go back to the split pages step and correct this by hand. If
your scans never need to be split, you can disable splitting (set it to manual
for all pages).
The third step is deskew, that is, a small rotation of each page to make the
orientation completely upright. Note that deskewing is applied separately to
every page, also to every split page. In most cases ScanTailor will correctly
make the orientation of the text as horizontal as possible. In very rare cases
you will have to adjust the deskewing by hand.
The fourth step is select content. It selects the rectangle that seems to contain all the text on the page. In quite a few pages this rectangle will be too
small or too big! (This is because it is difficult for the computer to understand
automatically what the actual text is and what is some artifact of scanning,
like a dark shadow at the edge of the page.) So it is at this step that you certainly will have to look at every page and check that the rectangle is selected
correctly. More about this below.
The fifth step is page layout. This step is fully manually controlled by the
user; each pages content rectangle is aligned (if desired) with the content
rectangles of all other pages, margins are added, and the resulting rectangle
is prepared.
Since it is only at step 4 that problems are quite likely to appear while step
5 is completely manual, I propose to run all the steps 1-5 automatically as
the draft run. After the draft run, you will have to return to step 3 and flip
manually through all pages to check that all is well. If needed, you will be
able to return to any previous step for every page where that step produced
an incorrect result. As experience shows, a non-negligible amount of work is
needed only for step 4 at this point.
The last step is output. At this step, which is usually quite slow but does
not require any attention from you, ScanTailor will produce the resulting TIFF
files in the output directory. After this step, you should flip through the
final page images again, and check that everything is okay (especially if there
were any color illustrations, see below). If there are no color illustrations, the
output is usually fine without any further manual work.
It is important to understand that your original scanned TIFFs will never be
changed; ScanTailor will only produce some new TIFFs in a different directory,
and this will be done only at the last step (the output). These TIFFs will be
the result of the ScanTailor processing.

4.4 Correct errors after the draft run

While the batch run is executing, you see the big stop button in the middle
of the ScanTailor window. You can stop the automatic operation of ScanTailor
at any time, by pressing on this stop button. Or you can save the project
23

file, also at any time without stopping the automatic run. This will save the
information gathered up to that point. (What if the power is cut to your computer? Then you will be able to continue right from the point where you last
saved the project file.)
When the draft run is completed, ScanTailor will stop and return to the first
page (figure 8).
Now you need to click on step 4 select content.
You will see an image of the first page with a rectangle around the text; this
is the rectangle that ScanTailor automatically selected according to its algorithms (figure 9). You will be able to see right away whether ScanTailor was
correct. Maybe on some pages text will be visibly cut off, or not included in
the rectangle. In order to correct all this, you will now flip through all the
pages in your project and correct all such possible errors. You will also be
able to immediately see and correct problems created at any previous steps
(1-4), such as incorrect splitting of double pages.
In the page shown in figure 9, everything is okay, so you go to the next page.
To flip to the next page, press PageDown or W on the keyboard. To go to
the previous page, press PageUp or Q on the keyboard. (Or you can use
the mouse wheel in the right column with thumbnails and then click on the
thumbnails.)
Note the long horizontal button over the thumbnails; this is the scroll lock
button. If this button is pressed, the thumbnail column will always show the
page you are currently working on. Otherwise you can scroll away from your
currently active page, to look at some other thumbnails.
As you go through the pages or switch between different steps, you may have
to wait a little bit as the display updates. Eventually, as you go through all
the pages, you will probably find a page where there is some problem after
the draft run. There are five main types of problems to be corrected; most
frequently:
1. the content rectangle needs adjusting (some text is outside, or the rectangle is too big and includes some noise)
2. the page alignment needs adjusting (usually at the beginning or at the
end of a chapter, when most of the text is at the top of page or at the
bottom of page)
3. incorrect splitting (this may happen when the page contains complicated
tables and so was split when it shouldnt have been)
4. incorrect deskewing (usually this happens when the page contains no
text but only some large shapeless illustration).
5. the scan was done incorrectly (e.g. the page was not completely scanned)
Let us see how these problems can be corrected.

24

4.4.1 Adjusting the content rectangle

You can see in figure 10 that the content rectangle is too small; some part
of the text was not included. Drag the content rectangle by the mouse, until
it is correct. The content rectangle, as a rule, should not include any white
margins the content rectangle should fit snugly around the text on the
page. White margins will be added later automatically.
An exception to this rule is when the content is neither centered nor flushed
on the page: e.g. a dedication appearing somewhere towards the upper part
of the page. In this case, it is easiest to make the content rectangle a little
bigger, so that the default page alignment (which is flush top and center
horizontally) produces good results.
Note: if some page in the book is completely empty, you should make sure
that it has no content rectangle at all. If any content rectangle was selected,
you should right-click on the page and click remove content rectangle. This
will speed up processing (a perfectly white page with correct sizes will be
generated).
Tip: While going through page contents, you will have to switch frequently
between the windows select content and page layout. It is possible to
switch between them by keyboard shortcuts: press P for Page layout and
press S twice for Select content. (The first S will get you to Split pages.)
4.4.2 Adjusting the page alignment
The page alignment options (see figure 7) are first, the sizes of the margins,
and second, the alignment of the content rectangle. The default options are
fine for most cases. (Note that the page will not be aligned unless the check is
set on the align with other pages checkbox.)
Sometimes the beginning of a chapter has text that is flushed down on the
page. You will have to correct the page alignment manually to flush down
for these pages, while for most other pages, most probably, flush up is the
right alignment option.
Sometimes you have a page with only very little text, or text that is only at
the bottom of the page, or only at the top. You need to adjust the alignment
of the page, or adjust the content rectangle so that it is aligned properly. For
example, see figure 11.
In this case the default page alignment (which is flush to top, center horizontally) will produce undesirable results (see figure 11, right). You can make
this page centered, but this is also not quite what you need. The easiest
is to adjust the content rectangle so that it is larger and is aligned properly
when flushed up and centered horizontally. Then you click on page layout
and see something like figure 12 right.

25

4.4.3 Adjusting the page sizes

If you notice in the page layout preview that the final image has very wide
or tall white margins on all pages, it means that some page is too wide or too
tall, and ScanTailor has adjusted all pages so that they have the same pixel
sizes. This can be fixed in two ways: first, you can exclude some pages from
alignment by unclicking align with other pages. Second, you can make the
margins smaller (say 0 mm) if you know that a significant amount of margin
will be added by alignment anyway.
4.4.4 Adjusting the splitting
You can see in figure 13 that the page image does not contain a small part
of the text. This cannot be fixed by adjusting the content rectangle because
some text is entirely missing from the image. (Note that the content rectangle
is also not quite right: it should be made a little taller to include the bottom
part of the table frame. But here we discuss the problem that cannot be
corrected by changing the content rectangle at this point.)

Figure 13: The left part of the page is missing and cannot be included in the
content rectangle at all.
26

The problem is that some part of the text was cut away at the splitting step!
Click on split pages and you will see something like figure 14.

Figure 14: The splitting step shows the line of splitting. It was obviously
incorrect.
Clearly, you need to drag the line of splitting to the left. After dragging that
line, click again on select content. Now you will see a better content rectangle; still it needs to be adjusted a little, until you see something like figure 15.

27

Figure 15: Problem corrected.

Note that in this example the scan did not actually need to be split. If this
is the case with your scans, you can disable splitting entirely. To disable
splitting, go to Split pages and select Change..., then Mode: Manual and
Scope: All pages to apply this setting to all pages.
4.4.5 Adjusting the deskewing
This is a rare problem. If you see that the page image is still significantly
skewed at the select content step, you need to click on the deskew step
and drag the blue anchor point with the mouse until the page angle is better.
Then you have to click again on select content and adjust the rectangle if
necessary.
4.4.6 Replacing scans in the project
Finally, you might discover that you scanned some pages incorrectly (e.g. some
part of the page was off the scanner glass). Then you can rescan that page
28

and add the new TIFF file to the project. Right-click on the thumbnail of some
page; you will see a menu Insert before, Insert after, Remove. This allows
you to remove incorrect scans and insert new, corrected scanned pages into
the project (although this is done one page by one page, so if you want to add
a lot of pages, it is better to start a new project).
Notes about removing or adding scans:
When you remove pages from the project, the scans are not actually
removed from the disk. Also, you can remove only one page from a split
double-page scan, if necessary.
It is advisable not to remove any empty pages in the middle of the book,
because removing these pages will break the numeration of the pages.
Empty pages will take practically no space in the final file. However, it is
better to remove empty pages at the very beginning and at the very end
of the book.
When you add pages to the project, the new pages will have no processing
steps already run on them, while other pages might be already partially
processed. So, for instance, the new pages will appear to have the default
page layout settings, and you will have to run all the steps on them,
including select content and page layout.
When a page has been removed from project, and this page was part of
a two-page sheet, the settings for the other half of the sheet will be lost!
You will have to click on that page and run again the content selection
and page layout steps.

4.5 Final run and final check-up

After going through all the pages and correcting the layout errors, you need to
return to the first page and click on the last step, output. You will see, after
a somewhat longer waiting, a final version of the first page and the output
options (see figure 16). The best options are: 600dpi, black/white mode, and
slight despeckling (small broom), as is the default.

29

Figure 16: Output options at the last step.

You should check if the brightness of the final image is okay. If you see that
the final picture is too dark and has lots of black dots around the text, you
should move the slider towards the thinner setting and wait to see the new
page image. If some of your scans are darker than others, you should scroll to
them and click on their thumbnail; this will prepare the final image and you
can then check whether it is too dark. Same if your final images are too light.
Note: If you see that some options are incorrect (e.g. Black/white should
be changed to Mixed or Color) while ScanTailor is still calculating the image,
you should change the options without waiting. Changing an option while a
process is running will automatically restart the process for this page!
If your book has important greyscale or color illustrations on some pages, see
section 4.6 for information about processing those pages. The other pages,
which contain only black/white material, will still have to be processed as I
will now describe.
If your book has all black/white text or black/white diagrams and no greyscale
or color illustrations, you just need to adjust the brightness so that the image
is sufficiently sharp and not too thin, not too thick.
You can view the page images at higher zoom: zooming is done by mouse
scrolling. Scroll up for zooming in, scroll down for zooming out. Then you
can see how the letters actually will look in the final image. You should expect smooth letter shapes at this point (ScanTailor applies some smoothing
algorithms to the scans).
30

Note that you can apply the brightness settings to all pages at once, or only
to selected pages (select them with the mouse in the thumbnail column at
right), or only to pages after the current one. It is important to remember that
the settings you click on the output window (or anywhere else in ScanTailor)
are only for the current page unless you press Apply and select Apply to all
pages or something else.
You are basically almost done at this point! Click again on the first page
thumbnail, so that you see the first page, and then click on the play button
to the right of output. This will start the automatic processing of all pages.
This operation is the final run, which may take an hour or more (maybe
about 15 seconds per page).
After this operation is done, you can do a final check-up of the pages. If the
images for some pages are somehow still not correct, you can go back to any
step and re-do it.
If your pages are all black/white, the only possible problems are these:
Final image is too thin/too thick on some pages where the brightness
was for some reason different from that of all other pages.
Despeckling has removed some dots that are actually part of the text.
You can flip through the pages while viewing the despeckling results: click
on the despeckling tab in the output window. The red dots will show where
ScanTailor removed dots from the image. If you see that ScanTailor removed
dots that are not dirt but actually are points in the text, such as . . .
somewhere, you should use a different despeckling broom or disable despeckling altogether (or make the image thicker). Usually, ScanTailor will be
careful with despeckling, but there are some cases when despeckling needs
to be disabled for some or all pages.
Note: it is advisable to save your project often while you are working on it.
ScanTailor is a stable program, but Windows is not, so if your computer
crashes for any reason, you will be able to continue right where you last
saved.
When you are done, the final images are in the output directory as a bunch of
TIFF files. These files will be in 600dpi and black/white, so they will be much
smaller than your original greyscale scans. This concludes the processing of
scans; the next step would be converting these scans to DJVU, see section 5.

4.6 Working with picture zones

Now let us see what you need to do if your book does have some greyscale or
color illustrations.
If your book has a lot of colored text (e.g., all chapter headings and all column
titles on each page are in blue), you should consider not making them colored but reducing all text to black/white. The colored chapter headings are
31

not particularly useful; making them all black/white will not significantly decrease the usefulness of the book, but it will significantly decrease the amount
of work you will have to expend on the file, and the final file size will be maybe
half the size.
If there are some pages with important illustrations, you need to navigate
to these pages and click on output. You will get to these pages if you flip
through all the final images after the output run. Do not wait until the final
image is produced and immediately click on Mixed in the Mode box.
In the mixed mode, ScanTailor will try to detect automatically where the
greyscale or color illustration is located on the page. As an example, see
figure 17.

Figure 17: In the mixed mode, the illustration below is automatically detected as the picture zone and is shown to you in changing color when you
click on the Picture Zones tab. Note that the upper illustration is purely
black/white and was not selected as a picture zone.
You can also adjust the brightness of the final image in the mixed mode.
Sometimes ScanTailor guesses the picture zones somewhat incorrectly. Then
you can draw your own picture zones with the mouse.
A few words about editing the picture zones. You can add new picture zones
with boundaries made of straight lines. You cannot delete the automatically
32

found picture zone. But you can substract a picture zone from the zones
already present. To do that, right-click on some point inside the picture zone
and select properties. Then you can select subtract from all layers or
subtract from the auto-layer.
If the automatically selected picture zone is very irregularly shaped, and if
this is not right, perhaps the easiest thing to do is to draw a big picture zone
around the automatically selected zone and select subtract from auto-layer,
so that the automatic picture zone is effectively removed, and then to draw
your own picture zones and select add to auto-layer. What you add to autolayer will take precedence over what you subtract from auto-layer. If you
click subtract from all layers, this is the highest layer and will subtract
also from your added layers.
The other possibility is not to tinker with picture zones but encode everything
as color. (The color mode.) In that mode, it is advisable to check the boxes
white margins and adjust luminosity. If you use this mode, the entire
image will be saved as a picture zone. This will result in larger files, but
is entirely acceptable and perhaps necessary if you have very complicated
graphics that are not greyscale. Experiment and see what works best for your
scans.
In any case, you can immediately see what the output will be for each given
page. You will have to experiment until you find the right options. You can
then apply these options at once to a group of pages or to all pages, by selecting the pages in the thumbnail column and pressing Apply To and then To
selected pages.

5 Encoding scans into DJVU

Once the processing of raw scans is finished, you have in the output folder a
bunch of TIFF files which are (almost all) black/white at 600dpi. These TIFF
files will take typically between 50 and 200 KB per page instead of about 4
MB that greyscale files took. By now you should have checked these TIFF
files and made sure that the quality of the black/white images is good: the
letters are sharp, have smooth shapes, there is little or no dirt etc. To check
all that, you can view the TIFF files in a picture viewer (such as IrfanView) at
high zoom.
Still, 50 to 200 KB per page is far too much. The next step is to encode these
images to DJVU format; this will reduce their size dramatically, typically to
5-10 KB per page.
To make a good, well-optimized DJVU file, you need one of the two programs:
either DjvuSolo version 3.1 or Djvu Document Express (DDE) 4.x, 5.x, 6.x
or Djvu Document Express Enterprise (DEE) version 5.1 4.x, 5.x, 6.x.12 The
DDE and DEE programs are much faster than DjvuSolo, and DEE 5.1 can
12

There is also a free software package called djvulibre, but it cannot produce sufficiently
well compressed DJVU files.

33

be configured to run in batch mode. On the other hand, DjvuSolo is a small

and freely downloadable program that requires no setup. The results in terms
of DJVU file quality from DjvuSolo and from DDE/DEE are pretty much the
same if you set the options correctly.
There are two ways of making DJVU files: one is by hand, another by batch.
To make a DJVU file by hand, run DjvuSolo or DDE and click FileOpen
to open the first TIFF file. Then click EditInsert pages... and select all the
other TIFF files. Please note: a selection box may have a bug in that you select
many files by holding the Shift key and the mouse but they will be selected
in the inverse order in the box. (This is a bug in a Windows dialog box.) Look
at the text in the file name field and check that you are selecting the files in
the correct order!
After inserting the pages you need to Save as... and select the Bundled
format for DJVU and Bitonal option at 600dpi. You can also edit the file
documenttodjvu.conf in the profiles directory and set pages-per-dict=100 or
200. The more pages per dictionary, the slower is the compression process,
but the smaller the resulting file size.
Note that the Bitonal option (or profile) in the DJVU encoders is intended
for purely black/white scans, while Scanned option is intended for scans
that have some (not many) colors but no photographs. Use the Photo option
for photographs.
To make a DJVU file by batch, you need DEE 5.1.13 First you need to create
a special set of options (or custom profile) for the DJVU encoding job. Run
the Document Express Configuration Manager, choose the profile Bitonal
(600dpi) from the list of profiles, click Advanced settings, and you will see
the following dialog.
13

This is a rather large package; there exists a stripped-down version that takes only about
20MB on the hard disk.

34

Now choose the Text tab as shown above. In that tab, set Pages per dictionary = 1000 (if this consumes too much RAM on your computer, or if this is
too slow, set to 200 or 300 instead of 1000). Save the custom profile under
a new name, say Bitonal-1. Do the same for the Scanned (600dpi) profile if
you need to encode books with color drawings.
Now run the Document Express Workflow Manager. Load all the TIFF pages
into it. In the Job name field, write the name of the book if you want. Choose
the previously created custom profile in the list Raster profile.

35

Then click to the Output tab (the tabs are at the bottom of the window). In
the list Separate document(s) choose One document only. Tick the box
under Enable at far left. Wait until the encoding is finished. You can also
look at the Log tab to watch the progress. Thats all; the DJVU file is created.
Do not delete the TIFF files yet! You may need to encode again if the DJVU
file has some error. Also, the TIFF files are useful for OCR purposes (see
section 6).
The result of DJVU encoding is a multipage DJVU file containing the entire ebook. You should rename that file to something sensible; not just math1.djvu.
At the very least, the file name should contain the authors name, the title of
the book, the publication year, and/or the ISBN number if available. This is
just a little work, but it will be so much easier to share that file on the Internet
if its name is sensibly chosen.

6 Creating text layer with OCR

Compared with the trouble needed to scan and process the book into a DJVU
file, it is really peanuts to add OCR for it. An e-book with search is a lot easier
to use.
The search in DJVU files works only if the DJVU file has the so-called OCR
layer. This layer is basically just a list of words stored inside the DJVU file
in compressed form. You can create the OCR layer using two programs:
FineReader and DjvuOCR. You need FineReader version 7 or 8.14 It is okay to
use even a trial or unregistered or evaluation version that you can download
for free. The result of running FineReader will be a set of FineReader batch
files. The wonderful program DjvuOCR created by Gencho will read these files
directly, extract the OCR information, and insert it into DJVU files.
Suppose you have already created the DJVU file out of some TIFF files. Hopefully, you didnt delete the TIFF files. Load the TIFF files into a new batch
in FineReader (keep in mind the problem with selecting many files at once!).
Set the recognition language and press Read all. When the OCR process
is finished, click Save batch. It is not recommended to edit the OCR text.
Previous versions of DjvuOCR could not process FineReader batches if the
OCR text was edited. The most recent version DjvuOCR 2.2, can deal with
small edits. You should not rewrite large blocks of text; i.e. you should keep
many original symbols in their original positions if you edit. Also you should
not delete the end-of-line symbols, so that the number of lines in a paragraph
remains the same. But we recommend that you do not edit the OCR text at
all. After saving the FineReader batch, you can quit FineReader and run the
program DjvuOCR.
14

FineReader 9 is now available but it cannot add OCR to DJVU files, and there is no
DjvuOCR support for FR 9.

36

This program has several functions; for example, DjVu Decoder will produce
TIFF files out of DJVU in case you deleted your TIFF files, or if you are working
with somebody elses DJVU file. For now, you will use only the Manual mode
OCR manager. Click that, and you get the following window.

Select the directory where the FineReader batch is located in the FineReader
Project directory field. Output OCR text file will be the name of the new file;
it doesnt matter what that name is. Tick the Burn DJVU file box and select
the DJVU file below; it means that the OCR data will be inserted (burned)
into the DJVU file. Click Process, wait a few minutes, and thats all. Now
the DJVU file is full-text searchable!
37

7 Adding book covers and color plates

It is reasonably easy to add a simple book cover. Just scan the book cover in
300dpi color, or even in 200dpi. Slightly blur the image in a graphics editor.
Encode into DJVU using the profile Photo(300) or Scanned. The resulting
1-page DJVU file needs to be inserted at the beginning of the DJVU e-book
after all the other processing is finished. Usually the book cover should not
be larger than 20-30 KB. It is probably not necessary to spend a lot of effort
on making a great-looking book cover. Consider that the people who will read
your e-book will spend most of the time reading the text rather than looking
at the cover.
In the same way one can add color plates, that is, special pages that contain
only color illustrations. Scan them separately and insert into the finished
DJVU file after all other processing is done.
To insert or rearrange pages in a DJVU file, use DjvuSolo or DDE. Open the
DJVU file, and you will see the thumbnails of the pages in the left column. You
can simply drag the thumbnails to rearrange the pages; you can also Cut,
Copy, and Paste pages or groups of selected pages, or delete pages. Use
the menu EditInsert pages... to add more DJVU pages to an existing DJVU
file. You can insert single-page or multipage DJVU files anywhere (before or
after any page), as you need.

8 Adding hyperlinks and bookmarks

After finishing all the preceding work with the DJVU file (including OCR),
you can add some hyperlink navigation to it. There are two ways of adding
hyperlinks.
The first is to use the DjvuSolo or Djvu Editor programs and add hyperlinks by
hand. Usually, one adds hyperlinks to pages in the table of contents for easier
navigation. In DjvuSolo or Djvu Editor you can select any rectangular area on
any page and then insert a hyperlink to a different page of the DJVU file. The
user will go to this page when clicking anywhere in the area. Note that the
hyperlink will point to a page number, so adding hyperlinks has to be done
after any changes to the page order or after inserting any additional pages
into the DJVU file. So if you want you can sit and make some rectangular
areas into hyperlinks until you are blue in the face.

38

The second way to add hyperlinks is semi-automatic, using the program DJVU
Hyperlinks Editor.15 Run the program and you will see the following window.

First you need to specify options for the hyperlinks Then you need to specify
the page range (
) in which the table of contents is located in the
DJVU file. These are DJVU page numbers, which may be different from the
page numbers printed in the book and in the table of contents (e.g. because
there are some pages taken by the cover and by the front matter). To compensate for this, usually one needs to add a certain offset to the page number; for
15

This program has only the Russian-language interface.

39

instance, page 10 in the printed book may be actually page 11 in the DJVU
file because one page is taken by the cover.16 Then you need to enter the
corresponding offset into the box

(offset). Now that all options are

(which means Add). This will add a new

enterd, press the button
DJVU file to the list in the left panel; the current options will apply to that file.
You can now set different options and add a different file. Finally, press the
button
the DJVU files.

(create). This will insert the hyperlink information into all

Similarly, one can create hyperlinks in the subject index. One needs to select
. The default entry
a different entry in the drop box
as shown means Table of contents. Other entries mean that you want to
process the subject index. The same settings apply.
After finishing the processing, one should view the DJVU file and check that
the hyperlinks were added correctly. The program relies on the OCR text for
determining the page numbers for hyperlinks. So any errors in OCR may lead
to errors in the position or targeting of the hyperlinks.

Where to download software

Name of program
IrfanView 4.1
ScanTailor 0.9.8
ScanKromsator 5.9
DjvuSolo 3.1
Djvu Editor 4.x, 5.x, 6.x (DDE/DEE)
FineReader 7.x, 8.x
DjvuOCR 2.2
Djvu Hyperlinks Editor

Download site
Status
www.irfanview.com
free
scantailor.sf.net
free
www.djvu-soft.narod.ru
free
www.djvu-soft.narod.ru
free
www.djvu-soft.narod.ru nonfree
www.abbyy.com
trial
djvuocr.ucoz.ru
free
www.djvu-soft.narod.ru
free

Big thanks to monday2000 for creating the website djvu-soft.narod.ru

Note for Linux users: All the programs in this table work reasonably well
under the standard Windows emulator (wine). However, some programs (IrfanView, DDE/DEE, FineReader) may fail to install if you run setup.exe for
those programs. You need to get portable or installed versions of these
programs that do not require running an installer. ScanTailor has a native
Linux version that can be compiled from the sources.

16

This is the Russian convention where the page numbering starts right away from the first
page of the book. In the Western typography the front matter usually has separate roman
numbering, so typical offsets will be not 1 but between 10 and 20.

40

Figure 7: After clicking on page layout while on the first page. The big
question marks on the thumbnails mean that these pages have not yet had
this step (page layout) performed on them. The Alignment symbols mean
the centering or flush-centering of the page in various directions. Press on
them to see immediately what effect these options would have on the final
appearance of the page.

41

Figure 8: After the draft run you are again at the first page. The big question
marks on the thumbnails are gone.

42

Figure 9: After you click on select content you can inspect the content rectangle. In most cases (like on this page), the content is detected perfectly.

43

Figure 10: In this case the content rectangle is too small. You need to adjust
it by dragging with the mouse.

44

Figure 11: The content rectangle is correct but very small (see on left). The
default page alignment will flush this rectangle to the top of the page and
center it, which is not what is desired.

45

Figure 12: An enlarged content rectangle (left) produces good page layout
(right) like in the original printed page.

46

Index
A3 scanner, 12
color plates, 38
deskewing, 11
DJVU, 4, 33
dictionary, 34
OCR layer, 36
rearrange pages, 38
FineReader
problems, 5
illustrations, 4
IrfanView, 7
JPEG, 8
digital artifacts, 9
problems, 8
kromsating, 14
quality, 3, 4
ScanKromsator, 5, 13
cutters, 15
draft run, 14
main run, 18
picture zones, 18
scanning, 11, 12
disk space, 12
greyscale, 4
with digital camera, 5
ScanTailor, 5, 19
TIFF, 8
upsampling, 4, 18
using Linux, 40
VueScan, 9

47