Due to the cost-effective advantage, range-free localization schemes are attractive for low-cost Internet of Things applications. As a distinguishing range-free scheme, DV-Hop localization can localize those unknown nodes which have less than 3 or even no neighbor anchors. However, the localization results by existing DV-Hop based algorithm are found to be inconsistent with the real connectivity between nodes. The inconsistency inspires us to propose two algorithms to improve localization accuracy. First a Centralized Connectivity-based DVHop (CCDV-Hop) algorithm is proposed to optimize the accuracy of DV-Hop localization. Establishing an optimization problem which takes the real connectivity between any two nodes as the constraints, the proposed algorithm can make the localization results conform to the real connectivity. Then an algorithm with lower complexity is proposed, namely Distributed Connectivitybased DV-Hop (DCDV-Hop) algorithm, which can obtain nearoptimal localization performance in distributed networks. Without including the connectivity of all nodes, the constraints in the proposed DCDV-Hop algorithm only consider the real connectivity within two hops. Simulation results show that despite higher complexity, the proposed algorithms can achieve much better accuracy than other DV-Hop based methods.