The recent advances in positioning technology facilitate the acquisition of high frequency and accuracy locational information of moving objects. Meanwhile, the locational information can be used to augment data collected with other sensors. One of them is to collect geo-referenced images in a road network and replay them continuously in a virtual-reality fashion to make users feel as if they are driving a car to explore the network in a virtual tour. With this background, an image-based road network model is defined. Trajectories and images collected with high-accuracy positioning devices and digital video camera, respectively, compose the model's input. Two algorithms are proposed to process the input and complete the model. With this model as the fundamental part, a prototype is developed to implement a virtual tour in a road network. In the virtual tour, instead of following a fixed route, users may decide the route choice and speed freely. Different from most existing methods, with the favour of advanced positioning technologies, the proposed approach and algorithms can facilitate building the image-based network model as well as developing the prototype efficiently and conveniently.