Abstract: This paper proposes an information theory model for progressive selection of streets.This model has two key components: ① information sources which represents the information of the street network.② channels which represents how to select the streets.A set of new quantitative measures is proposed for geometric information and topological information of street network.The geometric information considers the variation of length proportion occupied by the streets.The topological information considers variation of the connection relation of the streets in the street network.Channels for communication are built according to the idea of matching.The transition probability is the key issue.For geometric information channel,eliminated street matches to all the remaining streets.The transition probability can be calculated form the joint probability easily.For topological information channel,the connection relation matches to the connection relation still remained and to the connection relation disappeared after a street was eliminated.The non-connected relation matches to the non-connected relation before a street was eliminated.The map generalization meaning of the channels is explained.The experimental result shows that the eliminated streets are those of shorter and with lower degree.