Thumbnails are widely used in image retrieval and browsing systems. The use of such images compacts information. However, in situations where there are a large number of images on a server site, the use of thumbnails may still result in unacceptable response times when a user has access to a limited bandwidth. In this paper a refinement technique is proposed as an alternative to using thumbnails. The technique allows a user to identify the contents of an image early in its transmission. An efficient means of constructing a chain of image refinements of increasing quality to a given image is presented. Approximation chains can be employed either statically or dynamically. In the first case the level of detail is driven by the user; in the second case the level of detail is driven by a bandwidth aware server. Thus the proposed approach has application in grid and web computing.