Remote navigation systems, such as exploration of virtual 3D worlds or remote surgery, usually require higher bandwidth than the Internet connection commonly available at home. In this paper,we explore a set of techniques to reduce the bandwidth required by these applications. Our system consists of a client PC equipped with a graphics card, and a remote high-end server. The server hosts the remote environment and does the actual rendering of the scenes for several clients, and the new image is passed to them. This scheme is suitable when the data has a copyright or when its size may exceed the rendering capabilities of the client. The general scheme is the following: each time the position changes, the newviewis predicted by both the client and the server and the difference information between the predicted view and the correct one is sent to the client. To reduce bandwidth we can improve the prediction method, and the transmission system. We present here two groups of techniques: First, a set of lossless methods which achieve reductions of up to a 9:1 ratio. These are a combination of a two-level forward warping, that takes advantage of spatial coherence, and a masking method, which allows to transmit only the information that really needs to be updated. Second, a set of lossy methods suitable for very low bandwidth environments which involve both progressive transmission and image reuse. They consider relevant parameters such as the number of pixels, the amount of information they provide, and their colour deviation in order to create a strategy for prioritizing the information transmission. This system allows to improve up to an additional 4:1 ratio. The quality of the generated images is very high, and often indistinguishable from the correct ones.