A Novel approach for Denoising palm leaf manuscripts using Image Gradient approximations

This paper developed a novel algorithm to denoise old palm leaf manuscript using local Otsu thresholding and Sobel operator-based image gradient approximations. Initially an intense literature review is conducted on the existing document noise removal algorithms and implemented some of the algorithms which promised to give good PSNR values. Based on the results obtained, background noise removal algorithm using local features, gave the best results for denoising palm leaf manuscripts in terms of Peak signal to noise ratio and Mean Square Error. The values obtained are 14.64 db, 0.002 and 0.03 respectively. But the performance of this algorithm heavily depends on the threshold value used for filtering the document and is having a high computational complexity and overhead. So novel denoising algorithm is developed which can solve the drawbacks of the existing, document noise removal using local features algorithm. Using this enhanced algorithm, the PSNR, SSIM and MSE values obtained are 14.69, 0.003 and 0.02 respectively. So based on these parameter values, this novel approach proves best for palm leaves manuscript as compared to the various other algorithms existing for this purpose.