Tsinghua University at TRECVID 2004: Shot Boundary Detection and High-Level Feature Extraction.
Jinhui Yuan, Wujie Zheng, Le Chen, Dayong Ding, Dong Wang, Zijian Tong, Huiyi Wang, Jun Wu, Jianmin Li, Fuzong Lin, Bo Zhang · 2004
Our shot boundary detection system mainly consists of three components: fade out/in(FOI)detector, cut(CUT)detector and gradual transition(GT)detector. The key technique of FOI detector is the recognition of monochrome frame. In CUT detector, second order derivative method is applied to get cut candidates,and then flashlight detector and other gradual transition filter module are incorporated to eliminate false positives. In GT detector, short GT and long GT are treated separately, more specifically,twin comparison algorithm for short GT and finite state automaton (FSA) model for long GT. Ten runs are submitted. And the evaluation result shows that our system is among the best. The characteristics of the runs are summarized as follows. thuai02 Baseline system with default architecture and default parameter settings. thuai05 Second order derivative scheme for short GT. thuai07 Looser conditions for short GT detector. thuai08 Without gradual transition filter module. thuai10 Prolong the duration of post-cut modules,including flashlight detector and gradual transition detector. thuai14 Baseline system Derived from thuai10 with higher coefficients for motion based self-adaptive threshold. thuai15 Even higher coefficients for motion based self-adaptive threshold.