Techniques for supporting efficient content-based retrieval in multimedia databases

This paper presents a brief survey of the current state of the technology used in supporting efficient content-based retrieval in index-based multimedia databases that use feature vectors to represent multimedia data. One of the main obstacles in performing efficient content-based multimedia retrieval is the high-dimensionality of the feature vectors used to characterise the data. Previous surveys on multidimensional access structures tried to cope with this diversity bur tended to ignore the issues related to the high dimensionality of the feature space, This survey strives to give an overview of the techniques available to deal with such high dimensional spaces. The key properties of the most often used dimension reduction techniques will be compared, followed by a discussion of the multidimensional access structures that can deal with a high dimensional space. Algorithms for supporting content-based query processing using these multidimensional access structures will also be evaluated.