Readability enhancement of historical document images

CD-Rom included ; A Dissertation submitted to the Department of Computer Science and Engineering for the MSc in Computer ScienceThe Deportment of National Archives of Sri Lanka posses a huge collection of historical documents and those can be considered a national heritage as ic provides documental evidences for country's history. The documents any way are subject to many degradations due to natural aging, acidification, manual handling and lack of maintenance The department in the process of digitizing the documents as a preservation strategy and faced the problem of illegible document images after the digitization process. This research is conducted to find the solutions for this problem and particularly paid attention on ink-Weed-through problem, which occurs in aged, double sided documents.// We have evaluated many traditional image enhancement techniques like contrast stretching. histogram equalization and also paid attention on some other novice technologies like t-Means clustering and Principle Component Analysis. In order to conduct the experiments, we have used Image Processing Toolkit of Mat lab and written an application lo explore the performance of algorithms in depth. Further we have extended our experiments even lo many different color spares and results were analyzed in detail. This flexible approach has telped us to practically evaluate an existing algorithm with actual document images and 10 understand the pro? and cons of each approach.// We found that some traditional image enhancement techniques are contributing positively on historical document image enhancement, while some others are not. The application of commonly used image registration based techniques are found not feasible in real-time environment and certain other techniques like color segmentation seems capable of providing a reasonable output within a very short time AMeans clustering coupled with Principle Component Analysis is yet anther combined solution and its performance and the limitations are described .Based on the analysis, we have reached certain conclusions and scene recommendation*were made for a real-time historical document image enhancement strategy