Evaluation of Storage Structures in Image Database System

This paper addresses design criteria for the storage structure in an image database system that supports image access by area. These differ from convenitional criteria in that they are used when it is impossible to predict which area is accessed. Storage structures and access methods suitable for employing multi-page disk accessing are proposed and mathematically analyzed in terms of the expected I/O cost. Results show that (1) the optimum storage structures vary according to the size of image and that of the area required by users and (2) there exists an optimum page size, which is a function of the area size. Furthermore, if the area size is limited to a certain range, more effective solutions are applicable. It is expected that these findings will make it easier for designers of dafabase systems to select storage structures that meet their particular needs with regard to response time, image size, and area size.