This document describes research on developing a Khmer optical character recognition (OCR) system that is independent of font and size. It involves collaboration with another researcher and uses training data to recognize characters through preprocessing, feature extraction, and a recognition process using a Generic Fourier Descriptor. The goal is to produce a reliable OCR system for the Khmer language.
Best VIP Call Girls Noida Sector 22 Call Me: 8448380779
Khmer OCR System Overview Using GFD
1. Sovann EN
5th
Year Engineering student
Dept. Computer Science & Communication
Institute of Technology of Cambodia
Phnom Penh, Cambodia
Khmer OCR System
1Khmer OCR System
2. Native of research work
• A collaboration work with Mr. Kruy Vanna, PhDA collaboration work with Mr. Kruy Vanna, PhD
student at kameyama Laboratory , GITS, Wasedastudent at kameyama Laboratory , GITS, Waseda
UniversityUniversity
• The Objective is to produce a reliable Khmer OCRThe Objective is to produce a reliable Khmer OCR
system which is independent of Font and Sizesystem which is independent of Font and Size
2Khmer OCR System
3. Outline
• Overview of OCROverview of OCR
• Training dataTraining data
• Pre-processing and SegmentationPre-processing and Segmentation
• features extraction and recognition processfeatures extraction and recognition process
• Post-processingPost-processing
3Khmer OCR System
4. Outline
4Khmer OCR System
• Overview of OCROverview of OCR
• Training dataTraining data
• Pre-processing and features extractionPre-processing and features extraction
• Training process and recognition systemTraining process and recognition system
• Post-processingPost-processing
5. What is OCR ???
• Optical Character Recognition (OCR) is theOptical Character Recognition (OCR) is the
mechanical or electronic translation of scannedmechanical or electronic translation of scanned
images of handwritten, typewritten or printed textimages of handwritten, typewritten or printed text
into machine-encoded textinto machine-encoded text
http://en.wikipedia.org/wiki/Optical_character_recognitionhttp://en.wikipedia.org/wiki/Optical_character_recognition
5Khmer OCR System
• a document , a scan process , ocr system and ana document , a scan process , ocr system and an
out put text. (put all these pictures here. Png &out put text. (put all these pictures here. Png &
.docx).docx)
7. Overview of OCR System
Khmer OCR System
7
Training
data
Recognition
system
(Knowledge)
Pre-
processing
Features
extraction
Input
pattern
Character
Reordering
Recognition
result
Training process
Recognition process Features
selection/reduction
8. Outline
• Overview of OCROverview of OCR
• Training dataTraining data
• Pre-processing and features extractionPre-processing and features extraction
• Training process and recognition systemTraining process and recognition system
• Post-processingPost-processing
8Khmer OCR System
9. Training Data
9Khmer OCR System
• To cover more font and size, it is necessary toTo cover more font and size, it is necessary to
have more training sample of different fonthave more training sample of different font
Train Computer to recognize each of them isTrain Computer to recognize each of them is
១១
10. Outline
• Overview of OCROverview of OCR
• Training dataTraining data
• Pre-processing and features extractionPre-processing and features extraction
• Training process and recognition systemTraining process and recognition system
• Post-processingPost-processing
10Khmer OCR System
11. Pre-processing and Segmentation
• Pre-processing aims to produce data that arePre-processing aims to produce data that are
easy for the OCR systems to operate accuratelyeasy for the OCR systems to operate accurately
• The main objectives of pre-processing are :The main objectives of pre-processing are :
• BinarizationBinarization
• Particle removalParticle removal
11Khmer OCR System
12. Binarization (Thresholding)
• Image linearization (thresholding) refers to the
conversion of a gray-scale image into a binary
image.
12Khmer OCR System
13. Particle removal
• Salt-and-pepper noise is a kind of noise which isSalt-and-pepper noise is a kind of noise which is
usually caused by small unnecessary dotsusually caused by small unnecessary dots
produced by either the scanner or the sourceproduced by either the scanner or the source
document itself.document itself.
13Khmer OCR System
14. Segmentation
• Segmentation aims to produce each componentSegmentation aims to produce each component
to be recognized by the system.to be recognized by the system.
• The process is to separate the text of a pageThe process is to separate the text of a page
into each separate line, then to separate eachinto each separate line, then to separate each
line into Vertical Component, and finally produceline into Vertical Component, and finally produce
each independent symbol.each independent symbol.
14Khmer OCR System
15. Segmentation
• Segmentation aims to produce each componentSegmentation aims to produce each component
to be recognized by the system.to be recognized by the system.
• The process is to separate the text of a pageThe process is to separate the text of a page
into each separate line, then to separate eachinto each separate line, then to separate each
line into Vertical Component, and finally produceline into Vertical Component, and finally produce
each independent symbol.each independent symbol.
15Khmer OCR System
17. Feature Extraction & Recognition
• In feature extraction stage, each character is
represented as a feature vector which becomes
its identity.
• The major goal of feature extraction is to
extract a set of features which maximizes the
recognition rate with the least amount of
elements.
17Khmer OCR System
18. Recognition Process
• GFD is derived by applying two-dimensionalGFD is derived by applying two-dimensional
Fourier transform on a polar-raster sampledFourier transform on a polar-raster sampled
shape image.shape image.
18Khmer OCR System
19. Generic Fourier Descriptor
19Khmer OCR System
well…this iswell…this is កកកក
GFD Feature vectorGFD Feature vector
well…this iswell…this is កកកក
21. Recognition Process
• The similarity between two shapes is measuredThe similarity between two shapes is measured
by the City-Block distance of the two featureby the City-Block distance of the two feature
vectors of the shape.vectors of the shape.
• The lower value means the more similar theThe lower value means the more similar the
shapes are.shapes are.
21Khmer OCR System
23. Experimental Result
23Khmer OCR System
precision
Recall
F-Mesure
• The test was conducted on a document with aThe test was conducted on a document with a
resolution of 300 dpi of … symbols.resolution of 300 dpi of … symbols.
24. Khmer OCR Using Generic Fourier Descriptor Back
Thank for your attention !!!
24Khmer OCR System
25. Reference
28
Khmer OCR Using Generic Fourier Descriptor Back
[1] V. Kruy. Preliminary Experiment on Khmer OCR. Kameyama Laboratory,
Waseda Univerisy, Japan.
[2] Thesis for master degree, Khmer OCR, Vanna Kruy.
[3] D. Zhang and G. Lu. Shape-based image retrieval using generic Fourier descriptor.
Gippsland School of Computing and Information
Technology. Monash University. Churchill, Victoria 3842, Australia.
[4] Thesis for Doctoral Degree, chapter 6: Generic Fourier Descriptor, Dengsheng
Zhang.
[5] J.C.Rupe. Vision-Based Hand Shape Identification for Sign Language
Recognition. Department of Computer Engineering Kate Gleason College of
Engineering Rochester Institute of Technology Rochester, NY.
25Khmer OCR System
26. Reference
29
Khmer OCR Using Generic Fourier Descriptor Back
[6] D. Dimov. A polar-Fourier-Wavelet’s Transform for Effective CBIR. 3rd
ADBIS
workshop on Data mining & Knowledge Discovery
[7] I. Lengieng, K. Sochenda and C. Sokhour. , Khmer OCR for Limon R1 Size 22
Report, PAN Localization Cambodia (PLC) of IDRC.er OCR
[8] A. Averbuch, R.R. Coifmany , D.L. Donohoz M. Eladx M. Israeli. Fast and
Accurate Polar Fourier Transform.
Department of Computer Science, Tel-Aviv University, Tel-Aviv 69978, Israel.
Department of mathematics, Yale University, New Haven CT 06520-8283 USA
Department of Statistics, Stanford University, Stanford 94305-9025 CA. USA.
26Khmer OCR System