A unified approach to shot change detection and camera motion characterization

Abstract

This paper describes an original approach which jointly addresses two fundamental issues of video partitioning which represent the early important stage of any content-based video indexing system. These two issues are the detection of shot changes, and the labeling of the shot con guration related to the camera movement in terms of static shot, panning, traveling, zooming,... They are both derived from the computation, at each time instant, of the dominant motion in the image represented by a 2D a ne model, and from the variation of the size of its associated support. The successive steps of the method rely on statistical techniques ensuring robustness and e ciency. In particular, it can cope with scenes containing moving objects. Results on a real documentary video are reported and validate the proposed approach. Unité de recherche INRIA Rennes IRISA, Campus universitaire de Beaulieu, 35042 RENNES Cedex (France) Téléphone : 02 99 84 71 00 -International : +33 2 99 84 71 00 R sum : Nous d crivons une approche traitant conjointement deux probl mes fondamentaux du partitionnement de vid o en plan, la d tection des changements de plan et l' tiquetage d'une con guration de plan reli un mouvement de cam ra en termes de plan xe, traveling, zoom... Il s'agit d' une premi re tape importante pour tout syst me d'indexation de vid o par le contenu. La m thode propos e exploite, pour les deux aspects du probl me, l'estimation du mouvement dominant dans l'image, repr sent par un mod le a ne 2D, et de l' volution temporelle du support associ . Les tapes successives de cette m thode reposent sur des techniques statistiques, qui en assurent la robustesse et l'e cacit . Mots-cl : D coupage en plan, Mouvement de cam ra, Indexation vid o

A review of real-time segmentation of uncompressed video sequences for content-based search and retrieval

“…Bouthemy et al [16] based their detection on a dominant multiresolution motion estimation. This estimation uses a global 2-D parametric model composed of 6 parameters.…”

“…Detection if: 16)). Considering equations (50) and (51), the overall complexity of the method is O(4N B + 0.5B).…”

“….95 Joint probability images [55] O(2P + N 2 + N ) Hidden Markov Model with 3 features [26] O(13P + 11N ) [20,16,93,60] O(20P) + 80.80 Motion vectors [8,75,58] O(75P) + 2312.73 Optical flow [28] O(105P) + Frequency domain correlation [68] O(75P) + 105.65 [33] O (4P + 3N ) 16.47…”

Automatic acquisition of grammatical types for nouns

“…We use the system of (Bouthemy et al, 1999) which computes the pan, tilt, and zoom motions using the parameters of a two-dimensional affine model fit to every pair of sequential frames in a video segment. The output of this system is then clustered into characteristic camera motions (e.g.…”

Broadcast news parsing using visual cues: a robust face detection approach

“…Thresholding of motion-compensated frame differences is employed for shot change detection; this is a simple but fast approach for detecting shot cuts. More complex techniques for detecting shot transitions like zoom, wipe and dissolve effects can also be applied [2]. Shot classification is performed by estimating the class (anchor, report, static or outdoors) that best describes a whole shot, i.e.…”