This survey studies and evaluates four image retrieval systems on Internet. They are QBIC, ImageMiner, Netra and Excalibur. Image retrieval on these systems are content-based and similarity-based. Commonly supported content attributes are color, texture and shape which are represented in feature vectors or grid elements. Either distance measure or correlation method is used in similarity matching. Query by graphical example is commonly used. Based on these study, an ideal image retrieval system should be able to support high-level object based queries, modify the retrieval results based on user feedback, provide high level visual thesaurus, offer multilingual keyword search, permit users to sketch an object that they want to retrieve, and allow users to provide their own thumbnail images that they want to match against database images.