Adaptive Summarization of Digital Video Data

Abstract

As multimedia applications are rapidly spreading at an ever-increasing rate, efficient and effective methodologies for organizing and manipulating these data become a necessity. One of the basic problems with such systems is finding efficient ways to summarize the huge amount of data involved. In this chapter, we start by defining the problem of key frames extraction, then review a number of proposed techniques to accomplish that task, showing their pros and cons. After that, we describe two adaptive algorithms proposed to effectively select key frames from segmented video shots where both apply a two-level adaptation mechanism. These algorithms constitute the second stage of a Video Content-based Retrieval (VCR) system that has been designed at Old Dominion University. The first adaptation level is based on the size of the input video file, while the second level is performed on a shot-by-shot basis to account for the fact that different shots have different levels of activity. Experimental results show the efficiency and robustness of the proposed algorithms in selecting the near-optimal set of key frames required to represent each shot.