Citation Context

...ocus on reviewing papers published in the last four years, as a supplement to previous surveys [39], [181] on video summarization. A. Key Frame-Based Static Abstracts In recent years, many approaches =-=[182]-=-, [189] have been proposed to organize extracted key frames into static video abstracts. These approaches include video table of contents [142], storyboard, and pictorial video summary. For instance, ...

Citation Context

... 1: Performance of various video summarization methods on OVP. Ours and its variants perform the best. Unsupervised methods Supervised subset selection DT STIMO VSUMM1 VSUMM2 DPP + Q/D Ours (seqDPP+) =-=[30]-=- [31] [24] [24] [14] Q/D LINEAR N.NETS F 57.6 63.4 70.3 68.2 70.8±0.3 68.5±0.3 75.5±0.4 77.7±0.4 P 67.7 60.3 70.6 73.1 71.5±0.4 66.9±0.4 77.5±0.5 75.0±0.5 R 53.2 72.2 75.8 69.1 74.5±0.3 75.8±0.5 78.4±...

Citation Context

...mmarization techniques that produce a collection of static video frames (also known as video abstract or storyboard). In literature, different static video summarization techniques have been proposed =-=[9, 18, 19, 20, 12, 8, 11, 10, 16]-=-, most of them based on clustering techniques. In this case, the idea is to produce the storyboard by clustering together similar frames and by showing a limited number of frames per cluster (in most ...

Citation Context

...orm of scene clutter. Scene clutter can give rise to false alarms about the presence of targets in a scene. The implementation that is discussed also utilizes scale-space [24] and Delaunay clustering =-=[35]-=-, which, together with MCMCDA, represent a unique approach to vehicle tracking. This thesis is organized into three main discussion chapters. First, a background chapter (Section 2), provides insights...

Citation Context

...ization has attracted considerable interest from researchers and as a result, various algorithms and techniques have been proposed in the literature, most of them based on clustering techniques ([8], =-=[9]-=-, [10], [11]). Comprehensive surveys of past video summarization results can be found in [1], [12], [13]. In the case of clustering-based techniques, the basic idea is to produce the summary by cluste...

Citation Context

...-simages, as represented by sets of features, must besnumerically represented by an appropriate distance metric.sIn this work, four techniques were used: a method based onsDelaunay Triangulation (DT) =-=[27]-=-, STIll and MOving VideosStoryboard (STIMO) [14], Video SUMMarization (VSUMM)s[8], and VIdeo Summarization for ONline applicationss(VISON) [3].sThe Delaunay Triangulation (DT) [27] has severalsphases....

Citation Context

...mage included in the summary after each iteration and no keyframe is removed. Thus, this method generates strictly scalable storyboards. 4.4 Clustering A common approach in summarization is clustering=-=[51,14,12,28,40]-=-. Images are grouped into a few clusters {C1, · · · , CK}, each of them Ck = {Ik1 , · · · , IkP } containing similar images. Then each cluster is represented by a single image. Thus, the whole video i...

Citation Context

...e one hand, increasing the salience of the images selected for inclusion in the surrogate by means of a good method of segmentation, also makes that surrogate more representative for the source video =-=[18, 13]-=-. On the other hand, improving the layout of the selected keyframes by adding clues to the relations between and importance of represented segments, increases the information value of the summary [3, ...

Citation Context

...7 60.44 99.72 7 89.11 72.43 29.89 99.85 8 76.56 91.06 84.77 99.71 9 93.49 91.88 20.40 99.72 10 39.04 35.16 8.47 99.56 Average: 73.57 67.26 44.44 99.73aunary Triangulation based algorithm proposed in (=-=Mundur et al., 2006-=-). In the later, the frame contents are represented as multidimensional point data and then, they are clusteredusingDelaunary Triangulation, while the Open Video storyboards are generated by Table 3 R...