In this study, an integrated approach to identifying, analyzing and eliminating craquelures in high resolution oil paintings in images is proposed. It starts by identifying the craquelure networks in the oil painting via utilizing morphological top-hat operator, and follows by running a newly-devised crack line tracing algorithm to obtain properties of all the individual crack line segment such as the length and orientation. Closed regions in the craquelure netowrks is subsequently searched to extract further information about the craquelure patterns. Then, an improved recursive median filters is proposed to carry out the virtual restoration of the oil painting. The study concludes by presenting the experimental results which demonstrate good performance on oil painting craquelure analysis and virtual restoration.