To search, Click below search items.


All Published Papers Search Service


Video Data Modelling To Support Hybrid Query


Lilly Suriani Affendey, Ali Mamat, Hamidah Ibrahim, Fatimah Ahmad


Vol. 7  No. 9  pp. 53-61


Information contained in unstructured video data needs to be extracted and must be appropriately modeled in order to support storage and content retrieval. A video data model should be expressive enough to capture several characteristics inherent to video. Previous video data models lack connection among the video structure, the semantic and the image contents. Hence the types of queries supported are limited to the data models used. This paper proposes a video data model that would allow users to formulate hybrid queries on different attributes of video. The video data model captures the hierarchical structure of video (sequence, scene, shot and key frame), as well as high-level concepts (object, activity, and event) and low-level visual features (colour, texture, shape and location). With this representation, queries for the content and/or the specific hierarchical structure using similarity-based matching of low-level visual features as well as exact matching of textual attributes are supported. Experiments to compare query formulation using single types against hybrid query showed that hybrid query gives more accurate results. Thus, strengthening the need for a video data model that supports queries on more than one type of video attributes.


Video data modelling, video representation, video query, video database, video features