Evaluation of Different Histogram Distances for Temporal Segmentation in Digital Videos of Football Matches from TV Broadcast

Abstract

This article gives a more robust justification for the use of the Bhattacharyya distance in the algorithm used by our Automated Sport Analysis System named ACE in the first of its three perception modules. Such first module consists in the temporal segmentation of television video broadcasts, aiming to break down the video into shots, delimited by scene boundaries. An evaluation of other seven histogram distances is proposed, thus in order to better justify the use of the Bhattacharyya distance in ACE. Results presented were obtained using matches of the FIFA World Cup 2014. The classification results obtained by using Bhattacharyya distance are indeed the highest in terms of the F1-score, this was the main contribution of this study as there is no previous evidence of Bhattacharyya distance being compared with other distances and hence being selected as the one that produces the best classification results for our application.