Compact Visual Data Representation for Multimedia Search and Analytics

Shiqi Wang, Xinfeng Zhang

Abstract: With the exponential growth of multimedia in various forms, the volume of acquired visual data has dramatically increased while their value intensity remains relatively low. This presents significant challenges in multimedia search and analytics. In this tutorial, we aim to introduce recent advances of compact visual data representation techniques that enable efficient, flexible, and reliable multimedia search and analytics. We will explore the shift from traditional visual information representation techniques, such as video coding, to biologically inspired information processing paradigms, like digital retina based coding and representation. We will also discuss the representation of point cloud data and Artificial Intelligence Generated Content (AIGC) data, which are becoming increasingly popular in modern machine vision technologies. Additionally, we will discuss the recent advances in quality assessment technologies for multimedia signals under various novel and challenging scenarios. Finally, we will introduce the recent standardization activities in media coding including Video Coding for Machine (VCM). This tutorial aims to stimulate fruitful discussions, encourage innovative research, and drive advancements in the field of semantic and visual communication, multimedia search, analytics, computing as well as generative AI.