Proposal of Feature Matching Technique Using Similarity Features Filtering for Image Alignment  
Author Thanh Trung Nguyen


Co-Author(s) Hong Anh Thi Pham; Toan Van Pham; Thanh Minh Ta; An Van Tran


Abstract Recently, stemming from the storage of raw data into database systems for the purpose of further analysis and statistics has made digitizing paper invoices become one of the important tasks in many traditional businesses. The task is to automatically extract information from user-provided invoice image such as customer’s name, customer’s address, detail of transaction ans so on, then to store to database. In general, this task consists of three phases: (1) align the image with the template, (2) detect and extract the information regions in the image then, (3) recognize the optical character with deep learning technique. Image alignment is the most important step in the pipeline and directly affects the result, the accuracy of next step. In this paper, we propose a new approach for feature matching method called similarity features filtering and some techniques applying on invoices image pre-processing to improve the image alignment accuracy. The experimental results show that our proposed approach can achieve better results than other feature-based methods.


Keywords Image alignment, similarity features filtering, feature matching, paper invoices image alignment
    Article #:  DSBFI19-69
Proceedings of ISSAT International Conference on Data Science in Business, Finance and Industry
July 3-5, 2019 - Da Nang, Vietnam