XML Based Approach for Object Oriented Medical Video Retrieval Using Neural Networks

摘要

Large amount of audio visual medical information available in internet gets doubled every year. Also, conventional methods for medical video retrieval from large database seem to be inefficient, thus demanding for better video retrieval systems. This paper proposes an integration of several features extracted from key frames in order to develop an efficient medical video retrieval system. Extracted information from the frames was saved as object oriented format in the database. Video feature database or Object Oriented (OO) Database is stored with the extracted features of key frame (KF) images that includes entropy, frequency etc. This database holds the XML files containing the attributes of the key frames that is been extracted from it. During query, similar retrieval of XML files is obtained from input query video and similarity measure is applied on both query and in database to retrieve the best matching frames using XML file and corresponding videos is presented as an output. Thus this method provides an efficient retrieval of medical video from the database when a query is been submitted at the front end. Also, retrieval processing is improved with the help of neural network. This is performed as a stepwise operation, initially two attributes had been selected to search the information in the database and after the retrieval of certain key frame images all the other features had been applied over the database using neural network. The simulation results proved that the video retrieval from database using our combined approach provided better results than other video retrieval systems. Thus this method proved to be an efficient retrieval method for medical videos and other related videos.