Page 19 -

Contents xvii

Merging Type 2 Responses 313
Merging Type 3 Responses 315
Bagging and Boosting 315
Bagging 315
Boosting 316
Website Files 317
References 318

Chapter 9 Symbol Recognition 321
The Problem 321
OCR on Simple Perfect Images 322
OCR on Scanned Images — Segmentation 326
Noise 327
Isolating Individual Glyphs 329
Matching Templates 333
Statistical Recognition 337
OCR on Fax Images — Printed Characters 339
Orientation — Skew Detection 340
The Use of Edges 345
Handprinted Characters 348
Properties of the Character Outline 349
Convex Deficiencies 353
Vector Templates 357
Neural Nets 363
A Simple Neural Net 364
A Backpropagation Net for Digit Recognition 368
The Use of Multiple Classifiers 372
Merging Multiple Methods 372
Results From the Multiple Classifier 375
Printed Music Recognition — A Study 375
Staff Lines 376
Segmentation 378
Music Symbol Recognition 381
Source Code for Neural Net Recognition System 383
Website Files 390
References 392

Chapter 10 Content-Based Search — Finding Images by Example 395
Searching Images 395
Maintaining Collections of Images 396
Features for Query by Example 399
Color Image Features 399
Mean Color 400
Color Quad Tree 400