Ancient document compression and archiving via page decomposition

In this work we address the problem of document image compression, which has recently received increasing attention in the literature. To guarantee a high compression ratio and a satisfactory visual quality, the problem at hand requires sophisticated compression techniques, together with careful image preprocessing to remove noise, paper degradation artifacts, and ghost images from the back page. The strategy of the proposed technique is to segment the text from the background and apply the most convenient compression technique to the two resulting images. In particular, a pattern matching based compression procedure is applied to the binarized text image, while a lossy compression method is used for the background. The resulting image is perceptually almost lossless in its text component, while the background quality can be tailored to the desired compression ratio.