Using Exhaustive Threshold Approach for Identification of foreground and background images from non-printed part of Medical Prescription Document

Abstract

For processing of medical prescription document, foreground image and background image should be separated from non-printed part of medical prescription document. Different thresholding methods are used for this purpose. In this paper we investigate three methods (Global threshold, Adaptive threshold and Otsu’s threshold). For testing the performance of these methods, there is no ground truth database available in the literature. Hence, we have collected real time medical prescriptions from General physician, Ophthalmologists, Orthopedic Doctors, ENT specialists, Gynecologists and Dentists. For experimental purpose, we have considered more than 100 sample medical prescriptions. It is concluded that foreground image and background image segmentation from non-printed part of medical prescription by using Otsu’s threshold gives 71% result for character sharpness parameter and Global threshold provides 75% results for image brightness parameter.