Ranking highlight level of movie clips: A template based adaptive kernel SVM method. (April 2015)

• Record Type: Journal Article
• Authors: Wang, Zheng; Ren, Gaojun; Sun, Meijun; Ren, Jinchang; Jin, Jesse S.
• Is Part Of: Journal of visual languages & computing. Volume 27(2015)
• Page Start: 49
• Page End: 59
• Publication Date: 2015-04
• Subjects: Video analysis -- Highlight level -- Movie clip -- Template based method -- Adaptive kernel SVM; Visual programming languages (Computer science) -- Periodicals; Visual programming (Computer science) -- Periodicals; Programming languages (Electronic computers) -- Semantics -- Periodicals; Langages de programmation visuelle -- Périodiques; Programmation visuelle -- Périodiques; Langages de programmation -- Sémantique -- Périodiques; Programming languages (Electronic computers) -- Semantics; Visual programming (Computer science); Visual programming languages (Computer science); Periodicals; Electronic journals; 005
• DOI: 10.1016/j.jvlc.2014.10.015
• Languages: English
• ISSNs: 1045-926X

Abstract

Abstract: This paper looks into a new direction in movie clips analysis – model based ranking of highlight level. A movie clip, containing a short story, is composed of several continuous shots, which is much simpler than the whole movie. As a result, clip based analysis provides a feasible way for movie analysis and interpretation. In this paper, clip-based ranking of highlight level is proposed, where the challenging problem in detecting and recognizing events within clips is not required. Due to the lack of publicly available datasets, we firstly construct a database of movie clips, where each clip is associated with manually derived highlight level as ground truth. From each clip a number of effective visual cues are then extracted. To bridge the gap between low-level features and highlight level semantics, a holistic method of highlight ranking model is introduced. According to the distance between testing clips and selected templates, appropriate kernel function of support vector machine (SVM) is adaptively selected. Promising results are reported in automatic ranking of movie highlight levels. Highlights: A hybrid kernel based SVM model in predicting the highlight degree of clips. A set of new global and local features. A new clips database with users׳ marks.