1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
|
<page xmlns="http://projectmallard.org/1.0/"
type="topic"
id="finetuning">
<info>
<link type="guide" xref="index#configuration"/>
<link type="seealso" xref="manualeditionandcorrection"/>
<desc>Advanced options for a better recognition</desc>
</info>
<title>Fine-tuning</title>
<p><app>OCRFeeder</app> has some advanced options that can be
used to perform a better recognition. These options can be
chosen from the <guiseq><gui>Edit</gui><gui>Preferences</gui></guiseq>
dialog, under its <gui>Recognition</gui> tab.</p>
<p>The following list describes the mentioned options:</p>
<list>
<item><p><gui>Fix line breaks and hyphenization</gui>: OCR engines
usually read the text line by line and seperate each line with a
line break. Sometimes, this is not what the user wants because the
text might be broken in the middle of a sentence.</p>
<p>Checking this option will make <app>OCRFeeder</app> remove single
newline characters after the text is recognized by the engines.</p>
<p>Since just removing newlines in an hyphenized text would result
in wrongly separated words, hyphenization is also detected and removed
in this process.</p></item>
<item><p><gui>Window Size</gui>: <app>OCRFeeder</app>'s algorithm to
detect the contents in an image uses the concept of <em>window size</em>
which is the division of the image in small windows. A smaller window
size means it is likely to detect more content areas but size that is
too small may result in contents that should be part of a bigger area
instead. On the other hand, a bigger window size means less divisions
of contents but may end up in contents which should be subdivided.</p>
<p>A good window size should be slightly bigger than the text line spacing
in an image.</p><p>Users may want to manually set this value if automatic
one doesn't produce any valid content areas but normally it is easier to
use the automatic one and perform any needed corrections directly in
the content areas.</p></item>
<item><p><gui>Improve columns detection</gui>: Check this option if
<app>OCRFeeder</app> should try to divide the detected content areas
horizontally (originating more columns). The value that is used to
check the existance of blank space within the contents may be set to
automatic or manual when the columns aren't detected correctly.</p></item>
<item><p><gui>Adjust content areas' bounds</gui>: The detected content
areas sometimes have a considerable margin between their contents and
the areas' edges. By checking this option, <app>OCRFeeder</app> will
minimize those margins, adjusting the areas to its contents better.
Optionally, a manual value can be check to indicate the maximum value
of the adjusted margins.</p></item>
</list>
</page>
|
