Shot classification and scene segmentation based on MPEG compressed movie analysis

Authors:
Masaru Sugano;Masanori Furuya;Yasuyuki Nakajima;Hiromasa Yanagihara
Affiliations:
Multimedia Communications Laboratory, KDDI R&D Laboratories Inc., Saitama, Japan;Department of Electrical Engineering, Science University of Tokyo, Tokyo, Japan;Multimedia Communications Laboratory, KDDI R&D Laboratories Inc., Saitama, Japan;Multimedia Communications Laboratory, KDDI R&D Laboratories Inc., Saitama, Japan
Venue:
PCM'04 Proceedings of the 5th Pacific Rim conference on Advances in Multimedia Information Processing - Volume Part I
Year:
2004

Citing 8
Cited 0

Automatic recognition of film genres

Proceedings of the third ACM international conference on Multimedia
Automatically Segmenting Movies into Logical Story Units

VISUAL '99 Proceedings of the Third International Conference on Visual Information and Information Systems
Towards semantically meaningful feature spaces for the characterization of video content

ICIP '97 Proceedings of the 1997 International Conference on Image Processing (ICIP '97) 3-Volume Set-Volume 1 - Volume 1
Exploring Video Structure Beyond The Shots

ICMCS '98 Proceedings of the IEEE International Conference on Multimedia Computing and Systems
Linear Machine Decision Trees

Linear Machine Decision Trees
A fast audio classification from MPEG coded data

ICASSP '99 Proceedings of the Acoustics, Speech, and Signal Processing, 1999. on 1999 IEEE International Conference - Volume 06
Sound analysis using MPEG compressed audio

ICASSP '00 Proceedings of the Acoustics, Speech, and Signal Processing, 2000. on IEEE International Conference - Volume 02
Scene Determination Based on Video and Audio Features

ICMCS '99 Proceedings of the IEEE International Conference on Multimedia Computing and Systems - Volume 2

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper proposes shot classification and scene boundary/genre identification for MPEG compressed movies. Through statistical analysis of audio-visual features on compressed domain, the proposed method achieves subjectively accurate shot classification within the movies into a predefined genre set, as well as scene segmentation based on the shot classification results. By feeding subjectively evaluated feature vectors for each genre into the decision tree classifier, each shot is classified at very low computational cost. Then a sequence of shots belonging to the same genre is determined as a scene. The experimental results show that most of the shots in the movies are classified into subjectively accurate genres, and also that the scene segmentation results are more accurate and robust than the conventional approach.