Language Identification And Post Code Localization For Bangladesh Envelope Images

Automatic letter sorting is one of the applications of machine vision technology. It is a very important part of postal automation. It greatly affects whether or not the post office can realize automation in order to fulfill the increasing needs of letter processing. Automatic letter sorting depends on many machine vision and image processing technologies. These include image acquisition, document image analysis, layout analysis and pattern recognition.This thesis aims at practical application of the Bangladesh letter sorting system and proposes a segmentation technology for postal code of Bangladesh envelope images. Bangladesh is a multi-language country and it has two official languages --English and Bangladesh which have different characters and numerals, so it is necessary to identify the language of the envelope before postal code image segmentation and OCR reorganization. Based on this background and requirement, we present a novel and efficient technique for Bangla/English script identification with applications to the destination address block of Bangladesh envelope images.Firstly, we give the flowchart of the whole Bangladesh Letter Sorting System and the modules of Bangladesh Envelope Image Processing System. Secondly, we discuss the image processing technology; and present a postal code frame segmentation method based on the boundary check method. Thirdly, we extract the destination address block from the envelope image according to some page analysis methods and segment postal code from the destination address block. Finally, the advantages and disadvantages of current language identification methods are analyzed and a new method which can identify the language of Bangladesh envelopes effectively is proposed.