Journal of the Maharaja Sayajirao University of Baroda ISSN :0025-0422 Volume-55, No.1(VIII) 2021 94 AN EFFECTIVE METHOD FOR ANCIENT MARATHI SCRIPT READABILITY ENHANCEMENT Bapu Chendage & Rajivkumar Mente Department of Computer Applications, School of Computational Sciences, Punyashlok Ahilyadevi Holkar Solapur University, Solapur bdchendaage@sus.ac.in, rsmente@sus.ac.in Abstract: Ancient script contains large background noise. Most ancient images are with the same foreground and background. The background damage may happen due to varying contrast and duration of ancient images. The main aim of enhancement is to convert ancient images into readable form.Many techniques are available for image enhancement such as nonlinear multi-scale method, the image-enhancing method using thresholding, image enhancement using the Binarization and other methods. The paper describes image enhancement using Binarization and other methods for removing stains from unclear text script. These methods are tested on different types of scripts: script written on the stone, script written on metal plates, and document scripts.For proposed research work gives 55.7%, 62% and 65.6% accuracy using K-NN (K-Nearest Neighbor) classifier for the script on stone, metal plate and document type respectively. For the same images using SVM (Support Vector Machine) classifier the accuracy obtained is 53.2%, 59.5% and 67.8% respectively. Keywords: Marathi script, classifier, KNN, SVM, stain, readability, Background, RGB. 1. Introduction: Ancient script images play important role in epigraphy. This is a study of inscriptions on the stone, metal plate, and temple wall and pillar [3, 4]. Main problems in analyzing stone inscriptions are that they are not in readable form due to various natural climate conditions such as rain, wind, and light. And the location of the image also impacts the quality of ancient scripts. So many traditional methods are used for acquiring images like paper squeezes, rubbing, and scale drawing. The stains in color change in an ancient script. This stain may be created at the time of capturing or scanning the script image. For an effective preprocessing method ancient Marathi script images are converted into binary images and applied Thresholding to it. Image Binarization is used to translate pixels into two distinct groups white for the background and black for the foreground [4]. Thresholding plays a significant part in Binarization. The thresholding is of two type's local and global thresholding. In the case of Ancient Marathi script with uniform foreground and background, global thresholding is more suitable. And the image with variation in contrast and with large background noise, here foreground and background pixel cannot be removed, and in this case, local thresholding is used. The method includes calculating the local mean and standard deviation. 2. Local Enhancement using Mean and Standard Deviation: Mean is straightforward of all statistical measures. Means are frequently used in geometry and analysis. The complete illumination of the greyscale is measured using the mean. Standard Deviation is a generally used to calculate of variability in statistics. In terms of digital image processing, it displays how much variation exists from the mean value. A lowS.D shows that the data points tend to be very close to the mean, while a high Standard Deviation shows that the data points are extensive over a large range of values. The mean and standard deviation are calculated from the array of value using the following formula   =∑ n i = x i (1) n σ=√∑ n i=1 (x i -  ) 2 (2) n