Self-adaptive semantic data schema for multimedia databases

Abstract

In this dissertation, the target is to build up a multimedia database with query processing ability. A multimedia database should contain the image path attribute, location of the image from web or local computer, and the meta-data attributes, the image content descriptors. A user-friendly application program interface will be implemented for the user to make any image query from the well-developed database. In order to build up this advanced the multimedia database system, the ‘Separate Feature Databases Model’ and various algorithms of Content-based Image Retrieval (CBIR) are investigated. ‘Separate’ model requires a separate feature extraction pre-process for generating an additional feature database from the original image database while this is not required for ‘Direct’ Model. On the other hand, the computation requirement of ‘Direct’ model is much higher as all feature extraction and matching process need to be performed in the searching process. A various algorithms of Content-based Image Retrieval (CBIR) include the average RGB, RRG energy histogram, YUV energy histogram, FFT processing block and JPEG-7 edge detection energy histogram. Experimental systems have also been built and real life images in JPEG format are used for the experiments. The experimental systems are work on the different categories of database images, say flower, animate, human body and vehicles. Different algorithms will applied on different categories of images and obtain the top 5 results for comparison purpose. The medium multimedia database contains about 100 JPEG images is used.