A Generalized Zooming Technique for Pictorial Database System

Pictorial information processing relies heavily on the establishment of an efficient pictorial database system. Presentday database management systems are designed primarily for efficient storage, retrieval and manipulation of alphanumeric data. U ntiI very recently, little attention has been paid to the storage, retrieval and manipulation of non-alphanumeric information such as digitized images which require a large amount of storage even for pictures of average complexity. With the growing list of new applications in picture processing, such as geographic data processing, demographic data processing, computed tomography, whole-body scanner, earth resources survey satellite (LANDSAT) image processing, regional economic and health data processing, cartographic and mapping applications, etc., the problem of efficient, economical storage, retrieval and manipulation of vast amounts of pictorial information becomes more important and requires careful considerations. Two problems can be distinguished in designing pictorial databases-the storage, retrieval and manipulation of a large number of pictures, and the storage, retrieval and manipulation of pictures of great complexity. Traditionally, researchers in image processing have concentrated on working with a few pictures. However, the new applications for pictorial information sytems generally require that the systems be capable of handling a large number of pictures, some of which are also very complex. Consequently, new techniques must be investigated for the efficient, flexible retrieval of pictorial information from large pictorial databases. In Reference 2, an approach to designing an integrated database system for tabular data, graphical data, and image data is described. It is based upon generalizations of the relational approach to database design. 5 The main idea is to represent pictorial information by both logical pictures and physical pictures. A logical picture can be regarded as a model of the real image. It is defined as a hierarchicallystructured collection of picture objects. The logical picture can thus be stored as relational tables in a relational database and manipulated using a relational database manipulation language. Inquiries concerning the attributes of picture objects can also be handled by this relational database man-