Multiple inverted array structure for similar image retrieval

The retrieval functionality of a multimedia database system is important to provide users with comfortable access to the stored data. Similarity retrieval (also called content-based retrieval) is a promising technique for multimedia retrieval. We have therefore been carrying out research on an image retrieval system that makes use of similarity retrieval. This kind of retrieval is based on nearest neighbour searching in a multi-dimensional space. To make the retrieval fast and flexible we have developed a novel access method that uses an index structure consisting of multiple inverted arrays. Experimental evaluation of this method showed that it is potentially useful for shape similarity retrieval. The search speed is significantly faster than that of brute-force searching, and the accuracy is good enough