# Introduction n the era of modern technology, Optical character recognition has become one of the most significant topics. Optical Character Recognition is the Transformation of scanned images of printed, handwritten or typewritten text into machine-encoded text. In the natural process, human eyes are an optical mechanism. The image which is seen by the eyes is the input for the brain.OCR functions like the human ability of reading. OCR is a system prototype which has its scope, and it is using the Template Matching algorithm that is applied to recognize the character. In our paper, the key concept is OCR and Template Matching. Template Matching is the method which we use here for the proper implementation of OCR. # a) Basic Concept of Optical Character Recognition (OCR) OCR is the technique which performs automatic identification and allows to automatically recognize the characters through an optical mechanism. The goal of OCR is to classify visual patterns corresponding to alphanumeric or another letter.OCR is wanted when the information should be readable both to a human and to a machine. Alternative input cannot be predefined. # b) Basic Concept of Template Matching Template Matching is a technique for finding areas of an image that match a template. It is a high level machine visibility technique that destines the parts of an image that match a predefined template. It is flexible and relatively straight-forward to use which makes them one of the most well-known method. It is the process of finding the location of a subimage called template inside an image. Once several the corresponding templates is found, then the centers are used as similar points to determine the registration parameter. Template Matching involves determining the similarities between a given pattern and windows of the same size in an image and identifying the window that produces the highest similarities measure. It works by comparing derived image features of the image and the template for each possible displacement. # II. # Previous Work There are several approaches for text recognition and detection in images and videos have proposed earlier. These approaches can classify into two main classes: connected component based methods and texture based methods. A several numbers of research work on mobile OCR have founded. Motorola China Research center has presented a camera-based mobile OCR system for camera phones in [1]. An automatic text extraction system raised in [2]. An outline for a prototype Kanji OCR for recognizing machine printed Japanese texts and translated them into English is proposed in [3]. An approach of a character recognition system for Chinese script is in [4]. A system developed for only English capital letters in [5].A first skew correction technique for Camera Captured Business Card Images for Mobile Devices is proposed in [6].Optical Character Recognition remains very difficult for many languages. # III. # Implementation OCR takes input which is a text image and gives editable text document as output. The classification process of OCR is mainly two types. They are i) training and ii) testing. There are four steps in the OCR system. They are Pre-processing, Feature Extraction, Feature Training, and Feature Matching. # a) Pre-processing Here, the text image is a grayscale image which is converted into a binary image. A Binary image is worked with the pixel value 0, and 1. In the binary images, the letters constitute by binary 1(one) and the background constitute by binary 0(zero). # b) Feature Extraction The feature Extraction technique applies for all individual extracted character from the text image in the preprocessing steps. The objects which contain the pixel value fewer than 30 pixels will be removed. A flow diagram of the OCR system is given below: Here mainly two fonts namely Calibri (body) and Verdana have been considered for training data set. We will test the accuracy for other different fonts. # d) Feature Matching It is the comparison stage. When the feature value is matched with the trained feature, then the matched feature is set to recognize the exact character. The unmatched characters are considered as the defective characters. # e) Algorithm for Template Matching Method The main algorithm of this system prototype is the Template Matching algorithm. It is given below: STEP 1: The character image from the detected string is selected. # STEP 2: The image to the size of the first template is rescaled. STEP 3: After rescaling the image to the size of the first template (original) image, the matching metric is computed. # STEP 4: The highest match found is stored. # Experimental Result The experiment performed by giving any still images with fixed font size as input, and the following output is found. In the following experimental result, we can see that all the letters are visible. The Font named as Calibri and Verdana is considered as training data here. # Global Journal of Computer Science and Technology Volume XIX Issue II Version I V. # Discussion based on Experimental Result From above all the experimental result, we can reach a point of discussion. Although we have taken only two fixed sizes of fonts that is Calibri and Verdana as training data set, we have tested the accuracy for others five fonts namely Arial, Berlin Sans, Cambria, Lucida Fax and times new Roman. Experimental result on a set of images shows accuracy up-to 100%for Calibri, 100%for Verdana, 86.66% for Arial, 80%for Lucida Fax. Accuracy for Cambria and Times New Roman is very poor. We can improve the accuracy of these fonts by training the system with the character set of this fonts. character to produce text document. As an overall view of the system prototype, we can say that we develop this system prototype by using the technique that is the Template Matching approach to identify the character image. The interface of the system prototype looks very simple and makes the user of this system prototype easier to use it. For this reason, the recognition process of this system becomes user-friendly and smoothly because of the steps used in this system while recognizing the characters. # VI. Conclusions and Future Work In future work of this proposed algorithm can be enlarged and it will help the community in the field of handwritten character recognition. Except of alphabet and numbers, we can exalt the system with other characters. By ushering more features, we can enhance the accuracy. 1![Fig. 1: Flow diagram of the OCR system c) Feature TrainingHere mainly two fonts namely Calibri (body) and Verdana have been considered for training data set. We will test the accuracy for other different fonts.](image-2.png "Fig. 1 :") 5362![Fig. 2: Flow Chart for Template Matching IV.](image-3.png "STEP 5 : 3 STEP 6 :Fig. 2 :") 34![Fig. 3: Input Image](image-4.png "Fig. 3 :Fig. 4 :") 516273849![Fig. 5: Graph for Case-1](image-5.png "Fig. 5 : 1 Fig. 6 : 2 Fig. 7 : 3 Fig. 8 : 4 Fig. 9 :") ![There are several methods that have been introduced by different authors for optical character recognition. We serve a new technique to extract features from the images and identification of exactGlobal Journal of Computer Science and Technology Volume XIX Issue II Version I 34 Year 2 019](image-6.png "") 1Test ImageFont NameCorrect RecognitionIncorrect RecognitionAccuracyCase-1Calibri100100%(Consist of 10 characters)Verdana9190%Arial7370%Berlin Sans6460%Cambria4640%Lucida Fax7370%Times new Roman6460%Case-2Calibri200100%(Consist of 20 characters)Verdana16480%Arial17385%Berlin Sans14670%Cambria13765%Lucida Fax15575%Times new Roman71335%Case-3Calibri24196%(Consist of 25 characters)Verdana20580%Arial18772%Berlin Sans151060%Cambria141156%Lucida Fax18772%Times new Roman16964%Case-4Calibri300100%(Consist of 30 characters)Verdana25583.3%Arial26486.66%Berlin Sans191163.33%Cambria151550%Lucida Fax24680%Times new Roman171356.66%Case-5Calibri61198.38%(Consist of 62 characters)Verdana620100%Arial481477.41%Berlin Sans461674.19%Cambria352756.45%Lucida Fax402264.51%Times new Roman323051.61% © 2019 Global JournalsOptical Character Recognition based on Template Matching ( ) C © 2019 Global Journals Optical Character Recognition based on Template Matching * Design and implementation of a card reader based on build-in camera XLuo JLi LZhen International Conference on Pattern Recognition 2004 * Finding Textin Images VWu RManmatha EMRiseman Proc. of Second ACM International Conference on Digital Libraries ]K SBae KKKim YGChung WPYu of Second ACM International Conference on Digital LibrariesPhiladelphia, PA 1997 * Camera Based Kanji OCR for Mobile-phones: Practical Issues MKoga RMine TKameyama TTakahashi MYamazakiand TYamaguchi Proceedings of the Eighth International Conference on Document Analysis and Recognition the Eighth International Conference on Document Analysis and Recognition 2005 * Character Recognition System for Cellular Phone with Camera KSBae KKKim YGChung WPYu Proceeding of the 29th Annual International Computer Software and Applications Conference eeding of the 29th Annual International Computer Software and Applications Conference 2005 1 * A standalone OCR system for mobile camera-phones", Personal, Indoor and Mobile Radio Communications MLaine OSNevalainen IEEE 17th International Symposium 2006. Sept.2006 * A Fast Skew Correction Technique for Camera Captured Business Card Images AFMollah SBasu NDas RSarker MNasipuri MKundu Proc. Of IEEE INDICON-2009 Of IEEE INDICON-2009Gandhinagar, Gujrat 18-20 December * A Matlab Project in Optical Character Recognition (OCR) JesseHansen * Optical Character Recognition ThomasNatschlager Institute of Theoritical Computer Science * Dataset Generation for OCR" 10. Nadira Muda, NikKamariahNik Ismail AparnavaraLakshmivemuri TVSaikrishna Atulnegi Jasni Mohamad Zain "Optical Character Recognition By Using Template Matching (Alphabet) SitiAzami Abu Bakar * Optical Character Recognition LineEikvil 1993 * Evaluation of OCR Algorithms for Images with Different Spatial Resolutions and Noises QuinChen 2003 University of Ottawa * Optical Character Recognition Using Template Matching RachitVirendra Adhvaryu