Camera-based Document Recognition Using Hierarchical Classifier KYE

Abstract

This paper is to introduce a camera document recognition using hierarchical classifier. Generally, it is very difficult to recognize camera document image because of inconsistent input condition and camera problem itself. Therefore, Implementation of hierarchical classifier is needed to guarantee recognition performance of camera document image. The camera document goes through four processes such as preprocessing, segmentation and recognition modules. Input document image is evaluated by image enhancement algorithm and local threshold method. Characters are segmented using merging and split method, which is based on structural information of characters. Combined feature extractor and hierarchical MLPs have used to recognize a segmented character. We obtained an encouraging recognition result for camera document by combining image patches. We have experimented with ETRI document database and an encouraging recognition result 94.6% has obtained. Key-Words: Camera document recognition; Structural features; Hierarchical classifier

Topics

6 Figures and Tables

Download Full PDF Version (Non-Commercial Use)