Denoising, Analysis and Transmission of Manuscript Images

Abstract

The soul of image processing lies in denoising operation, text extraction and image compression. The main goal of image denoising is to enhance or restore a noisy image and help the other system to understand it better. The aim of image compression is to reduce the file size while keeping the image quality as high as possible, thus achieveing lesser storage requirement with smaller network bandwidth. In this thesis we perform denoising, text extraction and compression techniques on manuscript images.Manuscipt images are often corrupted with noise during acquision, transmission and retrieval from storage, which requires us to denoise them before further processing. Most of the manuscript images contain not only text and background regions, but also graphics. Therefore scanned manuscripts must often be segmented to make it ready for extracting whatever is desired. Finally these images have to be compressed for fast upload and download speeds and low storage requirement over internet. This thesis is divided into three major parts. In the first part of this thesis, we implement two different approaches for image denoising and compare their results for some internet images and manuscript images. The first approach is based on Markov Random Field (MRF) and the other is based on Non Local Mean algorithm introduced by A. Buades [1]. The experimental results show that Non Local Mean algorithm gives higher PSNR value and less visual artifact than other method based on MRF. In the second part of this thesis, we implement a text extraction method that uses edge information to extract textual areas from manuscript images, which is robust to color, effect of illumunation and the complexity of background. In the third part of this thesis, we implement a color image compression algorithm based on predictive based technique over the manuscript images. We apply it on original images, text extracted images and non text images which are the results of previous part and compare their results.