Certain Investigations On Restoration Of Degraded Ancient Handwritten Manuscripts

Abstract

A manuscript is a piece written by hand rather than newlinemechanically printed or replicated in some other indirect or automated newlinefashion. Historical manuscripts were written during the ancient period in newlinedifferent formats like paper, palm leaves, tree barks, wood, wall, stone, newlinecave walls, etc., which contain ancient information related to culture and newlinerich heritage. Most manuscripts are being conserved by various libraries, newlinemonasteries, and research institutions. newlineIn the olden days, writing in the Palm-Leaf was more popular newlinethan other media due to its easy availability and flexible storage facility. newlineThe quality of these old manuscripts is meagre due to their physical newlinedegradation and an assortment of other factors. Hence, a requisite is newlineintended for conversion into digital form, enabling future storage and newlinereference. newlineThis work aims to restore a good quality, neat, legible image newlinefrom the degraded historical document using prior knowledge and newlinethereby obtain a reasonable estimation of the required image as the newlineoutput. This task is done using various filtering methods and their newlinecombination with other reliable optimization algorithms. newlineThe thesis presents four enhancement techniques for Palm newlineLeaf Manuscripts (PLM), an ink bleed-through removal technique for newlinepaper manuscripts, and an effort on handwriting character recognition newlineby deep learning. newlinevi newlineThe first deals with denoising a given noisy PLM image of newlinedifferent dimensions subjected to various filters. An assessment is newlineperformed among the several statistical linear and non-linear filters with newlinevarying window sizes. The conclusion is made with the smallest newlinewindow size giving a superior performance where the images are newlinerestored using PM denoiser optimized using WOA to obtain the best newlinePSNR. newlineThe WOA-based binarization technique is employed to newlineoptimize adaptive thresholding. Here the WOA optimized adaptive newlineenhancement technique gave excellent results regarding various image newlinequality metrics like Accuracy, PSNR, DRD, etc. The whale optimization newlinestrategy has given good efficiency