TY  - BOOK
AU  - Lim,Wey Lin
TI  - Optical character recognition for business card
SN  - THE0002946(Local)
PY  - 2012///
CY  - Kuantan, Pahang
PB  - UMP
KW  - Image processing
KW  - Digital techniques
KW  - Optical pattern recognition
KW  - 	Computer vision
N1  - Project paper (Bachelor of Computer Science (Graphics & Multimedia Technology) -- Universiti Malaysia Pahang - 2012; Bibliography : p. 62-65
N2  - Optical character recognition (OCR) for business cards is a technology that transforms texts in images of business cards into machine readable texts. Although OCR systems for business cards are available in the market, they are dependent on specific business card scanner. Most scanners enable only one card per scan. Usually, organization spends efforts to manually input the data shown on business card to database. Alternatively, the business cards are scanned directly and stored into the  database. This process is time consuming and error prone. In order to address these issues, this thesis reports the development of an OCR for business card. The system takes the input from scanned image which contains a maximum of 8 business cards. The image is cropped into 8 individual images, each of which contains only one business card. The system then performs OCR functions in the image. The image data of the business card are transformed into machine readable texts. One of the advantages of the OCR system for business card is it does not rely on specific business card scanner. This study focuses on Malaysian business cards with Malay and English languages. The format requirement of the business card is specified with 26 alphabets, numerical values and punctuations with all the characters must be aligned in same orientation. The word recognition process is achieved in two passes, using Static Character Classifier and Adaptive Classifier. Static Character Classifier is trained by 60160 training samples that consist of 8 fonts in a single size with 4  attributes (normal, bold, italic and bold italic). Characters in images that are identified as blobs are then classified by Static Character Classifier. The result of the classification is passed to a dictionary containing a list of frequently used English words. Results that match the dictionary are then passed to Adaptive Classifier. Information collected in Adaptive Classifier is then used to improve the word recognition accuracy in the second pass of word recognition process. The system is tested using 36 different business cards and it is found that the average word recognition accuracy is 77.47%. The system developed will benefit users to better organize the information of business cards
ER  -