Implementation of Average Filter and Median Filter for OCR Pre Processing of Incoming Letters Image

Abstract

Letters are one of the tools to communicate through writing and are useful for conveying information. STMIK STIKOM Indonesia campus (called STIKI Indonesia) receives letters from other institutions reaching an average of 50 letters a month, an automatic system is needed that can help manage the incoming letters and can determine the purpose of the letter. OCR (Optical Character Recognition) is one of the ways that can be used to find a text contained in an image, but to get good results in the OCR process requires a pre-process in order to reduce noise in letter images, combining average filters and median filters be an option because it succeeded in reducing noise and smoothing the image of the organization’s stamp and signature that piled on the name of the sender of the letter. The results obtained from OCR after going through pre-processing are an array of words from letters. The accuracy achieved based on the date of the letter, letter number, subject of the letter and the sender of the letters from 50 test images reached 90%.