document analysis and recognition cs 661. what is a document? a.a written or printed paper that...
Post on 21-Dec-2015
212 views
TRANSCRIPT
Document Analysis and Recognition
CS 661
What is a Document?
a. A written or printed paper that bears the original, official, or legal form of something and can be used to furnish decisive evidence or information.
b. Something, such as a recording or a photograph, that can be used to furnish evidence or information.
c. A writing that contains information.
d. Computer Science. A piece of work created with an application, as by a word processor.
e. Computer Science. A computer file that is not an executable file and contains data for use by applications
Document Image Analysis
• DIA is the theory and practice of recovering the symbol structures of digital images scanned from paper or produced by computer
• DIA is a subfield of Digital Image processing– Digital images of natural objects: X-rays, fingerprints, faces,
scenery, etc. are NOT part of DIA– Digital images of symbolic objects: Postal addresses, printed
articles, forms, music sheets, engineering drawings, topographic maps belong to DIA
– Source: Scanners, printers, fax machines, hand!– Incidental text: license plates, billboards, subtitles, in photos
and video– WWW ??
• DIA’s grand goal is take us to the land of paperless office
Paperless Office?• Traditional transmission and storage of information has
been by paper documents• Documents are increasingly originating on the computer• Documents printed for reading, dissemination, and markup• Paper in the office has increased!!• Goal: Deal with the flow of electronic and paper documents
in an efficient and integrated manner• Implication: Unlike computer media, paper documents
should be read by both the computer and people
Short Tour of DIA
• Field started before digital computers could represent information traditionally appeared on paper
• Patents on OCR for telegraph and reading machines for the blind filed in the 19th century and working models demonstrated in 1916
• OCR on specially designed fonts used in 1950s
• First postal address reader installed in 1965
• OCRs to read scanned pages came into their own in 1980s with the advent of the low cost microprocessors, bit-mapped displays, and scanners
• Large capacity storage devices have now ignited the field with the prospects of Digital Libraries
• Document imaging today is a billion dollar industry but document interpretation is only a small part of it
Document Image Analysis
Graphical ProcessingTextual Processing
Optical Character
Recognition
PageLayout
Analysis
LineProcessing
Region and Symbol
Processing
Text Skew, blocks,paragraphs
Lines, curves, corners
Filled regions
Current
• Processors getting faster
• Storage costs are down– Pictures are typically 512 x 512 pixels
– Speech signals are typically 256 sample points
– Business letters are typically 2550 x 3300 pixels at 300 dpi
– Eng drawings are typically 34000 x 44000 pixels at 1000 dpi
• Digital libraries need WWW interface
• Information retrieval and search
• OCR accuracy on the rise
• Contextual models improved
Data capture
Pixel-level processing
Feature-level processing
Text analysis & recognition Graphics analysis & recognition
Document page
107 pixels
7,500 character boxes, 15x20 pixels each
500 line and curve segments, 20 to 20,000 pixels each
10 filled regions 20x20 to 200x200 pixels each
7500x10 character features500x5 line and curve features
10x5 region features
Document Description
1,500 words, 10 paragraphs,
1 title, 2 subtitles, etc.
2 line diagrams, 1 company logo, etc.
300 dpi, 8.5x11 in
255 gray
X 3 color
2,550 x 3,300 pixels
Processing Text Graphics
Pixels PreprocessingRepresentation, Noise removal, binarization, skew, script id, font id
PreprocessingRepresentation, Noise removal, binarization, thinning, vectorization
Primitives Glyph RecognitionConnected components, strokes, punctuations, words
Primitive RecognitionStraight lines, curve segments, junctions, nodes, loops, characters
Structures Text RecognitionWord segmentation, text line reconstruction, table analysis, linguistics
Structure RecognitionText fields, legends, labels, dimensions, graphics symbols
Documents Page Layout AnalysisText versus non-text, physical component analysis, logical component analysis, functional component analysis, compression
InterpretationComponent recognition, connectivity analysis, CAD layer separation, Database attribute extraction, Compression
Corpus Information RetrievalDocument Classification, indexing, search, security, authentication, privacy
Database, CADValidation, search, update
Document Image Analysis
Type Example DIA Task Ancillary Data
Plain text narrative Moby Dick Extract word order English lexicon
Newspaper, magazine
NY Times, Vogue Separate and reassemble articles, pointers to illustrations
Publication specific format
Scholarly, technical text
IEEE PAMI Index, author, title, page, figs, table, footnotes, equations
Abbreviations, acronyms, units
Formal text Program listing, chess, bridge, recipe
Extract executable form Program, chess, bridge syntax
Letter, Envelope Recommendation Sender, date, subject, routing info Directories
Directory Telephone book Extract name phone pairs Previous edition
Structured List Table of Contents Recover hierarchy, cross-refs Previous edition
Business Forms Order, invoice Convert to XML, link to Database Database form
Engineering Drawing
Part drawing, isometric view
Convert to CAD format Part list, drawing standards
Schematic Diag Circuits Convert to CAD format Constraints
Map Street map Convert to GIS format GIS, other maps
Music score Moonlight Sonata Recover MIDI representation Music syntax
Table Stock quotes Construct model; header-entries Stock abbreviations
Document Taxonomy
Postal ExamplesMeter Mark
Sender’s Address
Delivery Address
Linear Code
Digital Post MarkEndorsem
entIn Case of Undeliverable as Addressed Return to Sender
Forms
Unconstrained Text
Graphics Documents
Personal DL
DAS 02, Princeton, NJ• OCR Features and Systems
– Degradation models, script ID, Bilingual OCR, Kannada OCR, Tamil OCR, mp versus hw checks, traffic ticket reading
• Handwriting Recognition– Stochastic models, holistic methods, Japanese OCR
• Classifiers and Learning– Multi-classifier systems
• Layout Analysis– Skew correction, geometric methods, test/graphics separation, logical
labeling
• Tables and Forms– Detecting tables in HTML documents, use of graph grammars, semantics
• Text Extraction• Indexing and Retrieval• Document Engineering• New Applications
– CAPTCHA, Tachograph chart system, accessing driving directions
ICDAR 03, Edinburgh, UK
• Multiple Classifiers• Postal Automation and Check Processing• Document Understanding• HMM Classifiers• Segmentation• Character Recognition• Graphics Recognition• Non-Latin Alphabets- Kanji/Chinese, Korean/Hangul,
Arabic/Indian• Web Documents, Video• Word Recognition• Image Processing• Writer Identification• Forms and Tables
CS 661 Class ScheduleWeek M W
AUG 25 Introduction NIH /PROJECTS/Other Apps
SEPT 1 X IMG PROC for DIA
SEPT 8 Doc Analysis; GRAPHICS DIG LIB, Indexing, Retrieval
SEPT 15 Q PR Statistical PR- Structural, Neural Nets
SEPT 22 OCR OCR, CAMERA based problems
SEPT 29 Word Recognition Online
OCT 6 Q Handwriting Reco Paradigms PROJECTS
OCT 13 HMM formulation HMM, Viterbi, Baum-Welch
OCT 20 HMM Examples WEB DIA
OCT 27 Q Multilingual ; Language Models MCS, Committee methods
NOV 3 POSTAL, Forms BANK CHECK
NOV 10 Annotations; Historical; Other Apps
IMG PROC, filters (Gabor), transforms (wavelets, cosine)
Nov 17 Q CAPTCHA, Security Biometrics, watermarking
Nov 24 PROJECTS PROJECTS
Grading
• Home Assignments and Quizzes:– 4 x 10 = 40 points
– schedule is tentative to preserve surprise element
– Based on class participation and paper handouts
• Midterm project – Demo: 10%
– Report: 15%
• Final project – Demo: 10%
– Report: 25%
References
• Handbook of Character Recognition and Document Image Analysis, H. Bunke and PSP Wang (editors), World Scientific Press
• Document Image Analysis, Gorman and Kasturi , IEEE Computer Society Press
• International Conference on Document Analysis and Recognition proceedings
• International Workshop on Document Analysis Systems proceedings
• Symposium on Document Image Understanding Technology