A Systematic Review of Existing Data Mining Approaches Envisioned for Knowledge Discovery from Multimedia

Abstract

The extensive use of multimedia technologies extended the applicability of information technology to a large extent which results enormous generation of complex multimedia contents over the internet. Therefore the number of multimedia contents available to the user is also exponentially increasing. In this digital era of the cloud-enabled Internet of Things (IoT), analysis of complex video and image data plays a crucial role.It aims to extract meaningful information as the distributed storages and processing elements within a bandwidth constraint network seek optimal solutions to increase the throughput along with an optimal trade-off between computational complexity and power consumption. However, due to complex characteristics of visual patterns and variations in video frames, it is not a trivial task to discover meaningful information and correlation. Hence, data mining has emerged as a field which has diverse aspects presently in extracting meaningful hidden patterns from the complex image and video data considering different pattern classification approach. The study mostly investigates the existing data-mining tools and their performance metric for the purpose of reviewing this research track.It also highlights the relationship between frequent patterns and discriminativefeatures associated with a video object. Finally, the study addresses the existing research issues to strengthen up the future direction of research towards video analytics and pattern recognition.