Automatic categorization of tax forms and component block decomposition
Issued Date
2007
Copyright Date
2007
Resource Type
Language
eng
File Type
application/pdf
No. of Pages/File Size
xv, 255 leaves : ill.
Access Rights
open access
Rights
ผลงานนี้เป็นลิขสิทธิ์ของมหาวิทยาลัยมหิดล ขอสงวนไว้สำหรับเพื่อการศึกษาเท่านั้น ต้องอ้างอิงแหล่งที่มา ห้ามดัดแปลงเนื้อหา และห้ามนำไปใช้เพื่อการค้า
Rights Holder(s)
Mahidol University
Bibliographic Citation
Research Project (M.Sc. (Computer Science))--Mahidol University, 2007
Suggested Citation
Benjawan Pisuthisombut Automatic categorization of tax forms and component block decomposition. Research Project (M.Sc. (Computer Science))--Mahidol University, 2007. Retrieved from: https://repository.li.mahidol.ac.th/handle/20.500.14594/97018
Title
Automatic categorization of tax forms and component block decomposition
Alternative Title(s)
การจำแนกประเภทและการแยกส่วนประกอบของแบบฟอร์มภาษีโดยอัตโนมัติ
Author(s)
Advisor(s)
Abstract
This paper investigates methods for classifying types of tax forms and decomposing component blocks of tax forms automatically. We considered three methods. In the first method, the input image was first separated into component blocks; then types of tax forms were identified by comparing the component blocks with component blocks of template images. In the second and third methods, a type of input tax form was identified first and then form models and registration techniques were used to decompose the component blocks of the input tax form. In the second method, the type of tax form was identified by matching the tax form type image of an input tax form image and a prototype tax form type image, using a correlation coefficient. The last method identified the type of input tax form by recognizing characters and digits on the top of an input tax form. Experiments were performed on 520 tax form images composed of 26 types, each with 20 images. The first method achieved a 61.15% average correct classification result but it could not extract all component blocks correctly. With regard to the second and third methods, the accuracy rates of tax form identification were 88.65% and 100%, respectively, and they could extract all component blocks on tax forms correctly by using form models for component block decomposition. The results here showed that the method of using the character recognition on the tax form type and the form model had potential to be applied to develop the system for classifying type of tax form images and decomposing component blocks of tax form images.
Description
Computer Science (Mahidol University 2007)
Degree Name
Master of Science
Degree Level
Master's degree
Degree Department
Faculty of Science
Degree Discipline
Computer Science
Degree Grantor(s)
Mahidol University