The growing television and advertising industries have led to the tremendous amount of television video commercial (TVC) data stored in many data repositories of many orgizations. TV broadcasting, advertising, production houses, and marketing research agencies are to name of a few. these companies constantly use information from TVC data for various purpose such as: business intelligence, advertisment tracking, copyright control, etc. the previous techniques to manage image database is not based on visual features but on textual annotation of images. image database is indexed by textual description, i.e. keywords, captions, time of creation, etc. to facilitate queries [1,2]. however, the fast growing TVC databases have made manual annotation of images become expensive tasks. with thw advance of video warehouse technology, it is expected that a large volume of TVC database can be orginazed and analyzed to produce previously hidden information for decision making process. althougly there are a vast number of published works in video mining, little has been said about video mining on TVC data warehouse. for that reason, this paper proposes a multidimensional database as the first step in TVC video mining. the difference of this multidimensional database design from other related works is the focus given to both key low-level visual features and metadata for TVC video mining.