Imaged Document Text Retrieval without OCR

1
Imaged Document Text Retrieval without OCR

• IEEE Trans. on PAMI vol.24, no.6
• June, 2002
• ??????

2
Outline

• Introduction
• HTD and VTD
• Class of Character Objects
• Similarity Measure of Documents
• Experimental Results
• Conclusions

3
Introduction

• Retrieval of Imaged Documents
• Process with OCR v.s. without OCR
• Language dependence v.s. language independence

4
Procedure

• Image Preprocessing
• Feature extraction of character objects
• Horizontal Traverse Density (HTD)
• Vertical Traverse Density (VTD)
• Clustering
• To Identify classes of character objects
• Document representation
• Hash Table
• N-Gram
• To construct indexes for imaged document
  retrieval

5
Features HTD and VTD
6
Class of Character Objects

• Unsupervise Clustering with HTD and VTD
• Distance measure of character objects

7
Distance Measure of Character Objects
8
Examples of Character Objects
9
Similarity Measure of Documents

• N-Gram Algorithm
• Cosine angle between two documents

10
Corpus

• UW1 database (600 dpi)

11
Experimental Results

• Corpus I
• E01-E26

12
Experimental Results

• Corpus II

13
Experimental Results
14
Experimental Results
15
Experimental Results
16
Experimental Results
17
Conclusion and Future Work

• A new method for image document retrieval without
  OCR
• Retrieval of language independence
• Improvement of robustness for different fonts and
  noisy documents