Video shot representation based on histograms

  • Authors:
  • Tamires Tessarolli de Souza;Rudinei Goularte

  • Affiliations:
  • Universidade de São Paulo, São Paulo - Brazil;Universidade de São Paulo, São Paulo - Brazil

  • Venue:
  • Proceedings of the 28th Annual ACM Symposium on Applied Computing
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

The Content Personalization area, in digital videos domain, makes use of segmentation techniques to facilitate the processing of large data amounts while performing services as content selection or recommendation. Current researches have been directed to scene segmentation, which is based on clustering related shots. In this context, it is relevant to properly characterize shots using descriptors, because a good shot descriptor has potential to contribute to the improving of scene segmentation techniques already developed and, consequently, improve Content Personalization services. Our work aims to evaluate descriptors based on histograms for shot characterization. We present comparative results of scene segmentation by shot clustering using two approaches for shot representation: key frame and average histogram. The experiments were performed in two video domains: movies and television news. The results show that there aren't significant advantages between the approaches investigated and that they don't present good performance in assisting to identify scenes, showing the need of more research on new methods on shot representation.