Gossip based algorithms for information dissemination have recently received significant attention for sensor and ad hoc network applications due to their simplicity and robustness. However, a common drawback of many gossip based protocols is the waste of energy in passing redundant information over the network. Thus gossip algorithms need to be re-engineered in order to be applicable for energy constrained networks. In this paper, we consider a scenario where each node in the network holds a piece of information (message) at the beginning, and the objective is to simultaneously disseminate all information (messages) among all nodes quickly and cheaply. To provide a practical solution to this problem for ad hoc and sensor networks, NBgossip algorithm is proposed, which is based on network coding and neighborhood gossip. In NBgossip, nodes do not simply forward messages they received, instead, the linear combinations of the messages are sent out. In addition, every node exchanges messages with its neighboring nodes only. Mathematical proof and simulation studies show that the proposed NBgossip terminates in the optimal O(n)-order rounds and outperforms the existing gossip based approaches in terms of energy incurred to spread all information.