Publication

Abstract : Recovering of text details from degraded images is a difficult task due to high inter or intra variations between the foreground text and the background. The goal of document image restoration is to recover an actual image from the degraded image. In the proposed work, document image restoration based on Background Estimation and Expectation-Maximization algorithm is implemented using Matlab 2008b. Expectation Maximization algorithm estimates the parameters of the Gaussian Mixture Model and then extract the text data from the background based on Maximum Likelihood Estimation. The method shows less sensitivity to background noise and more sensitivity to the text details. The proposed method is performed and evaluated on various types of degraded document images with considerable background noise and different illumination conditions. The results of the proposed method show better performance over similar works reported in the current literature. © Research India Publications.