Output Constraints in Multimedia Database Systems

Abstract

Zusammenfassung AbstractSemantic errors exist as long as data are managed. Traditional database systems try to prevent this errors by proposing integrity concepts for stored data. Integrity constraints are used to implement these integrity concepts. However, integrity constraints can only detect semantic errors in elementary data.Multimedia database systems manage elementary data as well as complex media data, like videos. Considering these media data we need a much wider consistency concept as traditional database systems provide. Especially, data output of media data must be taken into account. In contrast to alphanumeric data the semantics of media data can be falsified during data output if data quality or synchronization of data are not suitable. Thus, we need a concept for output constraints that allow for preventing semantic errors in case of data output. For integrating output constraints into a multimedia database system we have to consider modelling, representation and checking of output constraints.For modelling output constraints we have introduced a constraint language which uses the same principles as traditional constraint languages. Our constraint specification language must support temporal and spatial synchronization constraints. However, it is desired to support both kinds of synchronization in almost the same manner. Therefore, we use Allen-Relations for defining temporal synchronization constraints as well as for defining spatial synchronization constraints.We need a database internal representation of output constraints that makes efficient constraint checking possible. The Allen-Relations used in the constraint language cannot be checked efficiently. However, difference constraints are a class of constraints that allows an very efficient checking. Therefore, we use difference constraints as database internal representation of output constraints.As methods for checking consistency of output constraints we use an approach based on graph theory as well as an analytical approach. Both approaches require a constraint graph as data structure. For data output we need an output order that is adequate to the defined output constraints. This 'output schedule' can be produced based on the output constraints.With output constraints, proposed in this thesis, semantical correctness of media data considering the data output can be supported. Thus, the contribution of this work is an qualitative improvement of managing media data by database systems.