Contents -- Foreword -- Preface -- 1. Document Image Analysis using Markovian Models: Application to Historical Documents -- 1.1. Introduction -- 1.2. Hidden Markov Random Field Models -- 1.2.1. Theoretical foundations -- 1.2.1.1. Simulated annealing -- 1.2.1.2. Iterated Conditional Modes (ICM) -- 1.2.1.3. Highest Confidence First (HCF) algorithm -- 1.2.1.4. 2D Dynamic Programming -- 1.2.2. Application of MRF labelling to handwritten document segmentation -- 1.2.2.1. Probability densities -- 1.2.2.2. Clique potential functions -- 1.2.2.3. Observations -- 1.2.2.4. Decoding strategy -- 1.2.3. Results -- 1.2.3.1. Zone labelling -- 1.2.3.2. Text line labelling -- 1.2.3.3. Conclusion -- 1.3. Conditional Random Field Models -- 1.3.1. Proposed model -- 1.3.1.1. Feature functions -- 1.3.1.2. Model inference -- 1.3.1.3. Parameter learning -- 1.3.2. A two level CRF model -- 1.3.2.1. Observation features -- 1.3.2.2. Label features -- 1.3.2.3. Learning -- 1.3.3. Integrating more contextual information -- 1.3.3.1. Global feature function -- 1.3.3.2. Combination of the information sources -- 1.3.3.3. Linear combination of the information sources (impl.2) -- 1.3.3.4. Combination of the information sources using an MLP (impl.3) -- 1.3.4. Experiments and results -- 1.4. Conclusions and Outlook -- Acknowledgments -- References -- 2. Information Just-in-Time: Going Beyond the Myth of Paperlessness -- 2.1. Introduction -- 2.2. Information Just-in-Time -- 2.2.1. Personal Information Environment -- 2.2.2. Hot/Warm/Cold Documents -- 2.2.3. Proposed Approach -- 2.3. Digital Pen Solution -- 2.3.1. Anoto Functionality -- 2.3.2. Data Entry Applications -- 2.4. iJIT Collaboration Platform -- 2.4.1. On-Demand Printing -- 2.4.2. Hybrid Document Management System -- 2.4.3. Research Notebook Application - iJITNote -- 2.4.4. Future Directions -- 2.5. Conclusions.

Acknowledgments -- References -- 3. The Role of Document Image Analysis in Trustworthy Elections -- 3.1. Introduction -- 3.2. History -- 3.3. Problems with Current Voting Technologies -- 3.4. Experimental Approaches to Reliable Processing of Voting Records -- 3.4.1. Statistical distribution of mark sense errors -- 3.4.2. Unbiased context-free visual auditing based on ballot images -- 3.4.3. Homogenous class display -- 3.4.4. Unique identification of ballots -- 3.4.5. Error characteristics of DRE with VVPAT -- 3.4.6. Development of testing procedures for voting systems -- 3.4.7. Affordances for voters with disabilities -- 3.5. Some Related Efforts -- 3.6. Concluding Remarks -- References -- 4. Information Retrieval from Document Image Databases -- 4.1. Introduction -- 4.2. Related Work -- 4.3. Word Shape Coding -- 4.3.1. Word Shape Coding by Character Stroke Categorization -- 4.3.2. Word Shape Coding by Character Boundary Extrema -- 4.3.3. Word Shape Coding by Character Holes and Reservoirs -- 4.4. Document Image Retrieval -- 4.4.1. Document Vector Construction -- 4.4.2. Document Similarity Measurement -- 4.5. Discussions -- 4.5.1. Coding Ambiguity -- 4.5.2. Coding Robustness -- 4.5.3. Document Similarity Measurements -- 4.5.4. Coding Scheme Selection -- 4.6. Conclusion -- References -- 5. Indexing and Retrieval of Handwritten Documents -- 5.1. Introduction -- 5.2. Template-based keyword spotting -- 5.2.1. DTW based keyword spotting -- 5.2.2. GSC feature based keyword spotting -- 5.3. Template-free keyword spotting -- 5.3.1. Segmentation of word image into character images -- 5.3.2. Extracting Gabor features from character image -- 5.3.3. Character classification probability Pr(ci

6.2. Check Preprocessing -- 6.2.1. Check processing challenges -- 6.2.1.1. Legal amount recognition challenge -- 6.2.1.2. Check processing challenges -- 6.2.1.3. Date recognition challenge -- 6.2.1.4. Payee name recognition challenge -- 6.2.1.5. Signature detection challenge -- 6.2.1.6. Courtesy amount field -- 6.3. Legal Amount Recognition -- 6.3.1. Legal line parsing -- 6.3.1.1. From word to legal line phrase matching -- 6.3.2. Legal line recognition for handwritten checks -- 6.3.3. Handwritten word recognition -- 6.3.3.1. Slant correction -- 6.3.3.2. Character segmentation -- 6.3.3.3. Word matching algorithm -- 6.3.3.4. Legal line phrase matching -- 6.4. Experimental Results for CAR and LAR -- 6.5. Automatic Signature Verification -- 6.5.1. Feature-based ASV -- 6.5.1.1. Description of features -- 6.5.2. ASV based on dynamic feature matching -- 6.5.3. Multi-references ASV matching function -- 6.5.4. Results -- 6.6. Conclusion -- Acknowledgments -- References -- 7. Statistical Deformation Model for Handwritten Character Recognition -- 7.1. Introduction -- 7.1.1. Contributions of this Chapter -- 7.2. Statistical Deformation Model of Offline Handwritten Character Recognition -- 7.2.1. Extraction of Deformations by Elastic Matching -- 7.2.2. Estimations of Eigen-Deformations -- 7.2.3. Recognition with Eigen-Deformations (1) -- 7.2.4. Recognition with Eigen-Deformations (2) -- 7.2.5. Recognition Result -- 7.2.6. Related Work -- 7.3. Statistical Deformation Model of Online Handwritten Character Recognition -- 7.3.1. Extraction of Deformations by Elastic Matching -- 7.3.2. Estimation of Eigen-Deformations -- 7.3.3. Recognition with Eigen-Deformations -- 7.3.4. Recognition Results -- 7.3.5. Related Work -- 7.4. Conclusion -- Acknowledgment -- References -- 8. Robust Word Recognition for Museum Index Cards with the SNT-Grid -- 8.1. Introduction.

8.2. Synthetic Training Data -- 8.3. The Scanning N-Tuple Grid -- 8.3.1. Training and recognition of the SNT-Grid -- 8.3.2. Fast encoding with a two pass templates -- 8.4. Isolated Characters Results in Cards with Synthetic Data -- 8.5. Priority Queue Method -- 8.6. Experimental Results -- 8.6.1. Mis-recognized examples -- 8.6.2. Correctly recognized examples -- 8.7. Discussion and Conclusions -- References -- 9. Historical Handwritten Document Recognition -- 9.1. Introduction -- 9.2. Related Work on Historical Handwritten Document Recognition and Retrieval -- 9.3. Classification Models for Handwritten Word Recognition -- 9.3.1. Support Vector Machines -- 9.3.2. Conditional Maximum Entropy Models -- 9.3.2.1. Discrete Predicates -- 9.3.2.2. Continuous Predicates -- 9.3.3. Naive Bayes with Gaussian Kernel Density Estimate -- 9.4. Sequence Models for Word Recognition -- 9.4.1. Word Recognition with Discrete HMMs -- 9.4.1.1. Feature Probability Smoothing for HMMs -- 9.4.2. Conditional Random Fields Framework -- 9.4.2.1. Inference and Training in CRFs -- 9.4.2.2. Training and Inference with Beam Search -- 9.4.2.3. Word Recognition with CRFs -- 9.4.3. HMM with Gaussian Kernel Density Estimates -- 9.5. Experimental Results -- 9.5.1. Experimental Setup -- 9.5.2. Results on Different Classification Models -- 9.5.2.1. SVMs -- 9.5.2.2. Conditional Maximum Entropy Models -- 9.5.2.3. Naive Bayes with Gaussian Density Estimate -- 9.5.3. Tune and Compare Beam Search for Our CRF Model -- 9.5.4. Result Comparisons -- Acknowledgments -- References -- 10. Statistical Modeling of Document Appearance -- 10.1. Statistical appearance models -- 10.2. Layout grammars and parsing -- 10.2.1. Computational challenge -- 10.2.1.1. Exponential number of segmentations or groupings -- 10.2.1.2. Context sensitivity -- 10.2.2. Generative context free and attribute grammars.

10.2.3. Non-generative grammars -- 10.3. Probabilistic Markov grammars -- 10.3.1. Separable layouts and Document Image Decoding -- 10.3.2. Turbo recognition -- 10.4. Local feature based models -- 10.4.1. Features measured in subregions -- 10.4.2. Local descriptors -- 10.4.3. Locational distribution of intensity variations -- 10.4.4. Feature descriptors without location -- 10.5. Graph representations of layout -- 10.5.1. Hidden tree Markov model -- 10.6. Models of multi-line structure -- 10.6.1. Tables -- 10.6.2. Mathematical expressions -- 10.7. Models of textline structure -- 10.7.1. Baseline skew and warping -- 10.7.2. Glyph layout on baseline -- 10.7.3. HMMs for textlines -- 10.8. Models of character and stroke appearance -- 10.9. Research frontiers -- References -- 11. Reverse-Engineering of PDF Files -- 11.1. Introduction -- 11.2. Related Works -- 11.2.1. Related Tools -- 11.2.2. Related Researches -- 11.3. PDF File Format -- 11.3.1. Internal File Structure -- 11.3.2. Content Streams -- 11.3.2.1. Text Representation -- 11.3.2.2. Graphics Representation -- 11.3.2.3. Images Representation -- 11.3.3. Structuring the Content -- 11.4. Reverse Engineering Approach -- 11.4.1. Physical vs. Logical Structures -- 11.4.2. Global Architecture -- 11.5. XCDF Canonical Format -- 11.5.1. Requirements -- 11.5.2. Format Description -- 11.5.3. Representation of Text Blocks -- 11.5.4. Representation of Vector Graphics -- 11.5.5. Representation of Images -- 11.5.6. Space Coordinates and Device Independence -- 11.6. Document Analysis with XED -- 11.6.1. Parsing PDF Objects -- 11.6.2. Creating the Virtual Document -- 11.6.3. Layout Analysis -- 11.6.4. Evaluation of XED's Physical Structures Extraction -- 11.7. Document Understanding with Dolores -- 11.7.1. Logical Structures and Document Models -- 11.7.2. A First Step Toward Document Reconstruction.

11.7.3. Recovering the Logical Structure from Newspapers.