Structural isomorphism between languages benefits the performance of cross-lingual applications. We propose an automatic algorithm for cross-lingual similarization of dependency grammars, which automatically learns grammars with high cross-lingual similarity. The algorithm similarizes the annotation styles of the dependency grammars for two languages in the level of classification decisions, and gradually improves the cross-lingual similarity without losing linguistic knowledge resorting to iterative cross-lingual cooperative learning. The dependency grammars given by cross-lingual similarization have much higher cross-lingual similarity while maintaining non-triviality. As applications, the cross-lingually similarized grammars significantly improve the performance of dependency tree-based machine translation.