Query Processing in the MultiMedia Visual Information Seeking Environment: A Comparative Evaluation

Although much research has been conducted in the area of multidimensional range queries, we examine this problem from a new perspective. In particular, our goal is to support the processing of incremental multidimensional range queries specified via our temporal visual query language (TVQL), a direct manipulation visual query interface. In this paper, we describe the details and context of our problem, emphasizing the need to support temporal browsing and examining the characteristics of temporal data. We present a simple but new array-based index structure called our k-Array and its bucket-based counterpart, the k-Bucket. We then describe the series of experiments over several temporal queries that we ran to compare the k-Array and k-Bucket with basic methods such as the linked array and other popular bucket-based methods such as the grid file and k-d tree. Our results show that the k-Bucket performs the best overall, even though it is subject to projection effects. The other methods also perform fairly competitively, but only under certain conditions.