Vectorization of Glyphs and Their Representation in SVG for XML‑based Processing

Institut für Print- und Medientechnik der TU Chemnitz[Institute for Print and Media Technology • Chemnitz University of

Technology] Direktor: Prof. Dr. Arved C. Hübler • Reichenhainer Str. 70 • 09126 Chemnitz • Germany

http://www.tu-chemnitz.de/pm • [email protected] • Tel: +49-371-531-2364 • Fax: -3780

Vectorization of Glyphs and Their Representation in SVG for XML‑based Processing Stefan Pletschacher; Marcel Eckert; Arved C. Hübler

2 Pletschacher • Vectorization of Glyphs and Their Representation in SVG for XML‑based Processing • ELPUB 2006

Digitization of Historical Documents

GEB1150


Alphabet und Font Extraction

XML instance

alphabet and fontdefinition

...

content

...

glyph ID1

glyph ID2

ID1 ID2 ID3ID3 ID4


Vectorization - Raster to Vector Conversion

font assignmen

t

Vectorization

RIP

41 hex

OCR

vector font

encoded text e.g.

ASCII

bitmap graphic


document image

blocks

textual blocks image blocks

structural information

region basedsegmentation

blockclassification

text lines

segmentation

words

characters

segmentation

segmentation

DIA System und Workflow

1. text (headline)

2. bitmap image 3. text block

4. text block

…

1. text (headline)


4. text block

…


character images

set of prototypes

clustering

vectorisable glyphs

classification ofvectorisable glyphs

non vectorisable

images

set of bitmap symbols

IDassignment

set of vectorised

paths

vectorisation

document specific

SVG font

transformationto SVG

set of SVG glyph

descriptions

assignment of private Unicode code points


&#xE000



XML + SVG encoded

document

image blocks

structural information

set of bitmap symbols

document specific

SVG font

encoding

references

specific output formats

layout modificationby means of XSLT

OCR

XML

1. text (headline)


4. text block

…

1. text (headline)


4. text block

…


Vectorization Approaches

• Contour based

• Skeleton based

CompxNCompxCore )(: CorexCompxCont :

zxyxCompContxdist

zyCompContzyCompxS

)(,(

, ),(,:


Applied Algorithms

• Pre-processing- Finding connected components (Region Growing)- Contour extraction (Contour following)

• Polygonal Approximation Based on Relaxation- Phase 1: Clustering of polygonal points- Phase 2: Relaxation (Error correction)

• Automatic Parameter Control- Rasterization of the resulting glyph images- Ascertaining a weighted error (Ground Truth)- Selecting appropriate vectorization parameters


Finding Connected Components

Ü Ö Ä % “ !


Region Growing


Contour Following

white pixel

black pixel

starting point

examination order


Clustering of Polygonal Points


Relaxation


SVG Representation


Visual Quality


Formal Quality Measurement - Ground Truth

Error function- absolute number of wrong pixels- weighted by the distance to the next true component


Results


0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 1,1 1,2

vectorization parameter ε

acc

ura

cy

H K d

Adaptive Parameter Control

-5

-4

-3

-2

-1

0

1

2

0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 1,1 1,2

vectorization parameter ε

accu

racy

gra

die

nt

H K d


Compression rates


Conclusions

• Good vectorization results already with linear primitives• High compression rates can be achieved• Extracted fonts can be easily scaled and further formatted• Known vectorization methods have been extended towards an adaptive system for automatic parameter control• These methods can be applied for preservation and handling of unknown type faces in digitized documents• Originals may be re-encoded using a document specific alphabet and font• Direct integration into XML/SVG based processes possible• Various output formats can be supported by means of XSL transformations


Thank you very much!

[email protected]

Questions