A Study of Flash Scene Detection in Steaming Video

碩士 === 中原大學 === 資訊工程研究所 === 104 === Shot Boundary Detection is a fundamental step of streaming video analysis and indexing. In the past decade, a lot of researches have been proposed for shot boundary detection. However, these methods are not effective if flashlight appeared in the videos. In this p...

Full description

Bibliographic Details
Main Authors: Chun-Ching Lee, 李俊慶
Other Authors: Slo-Li Chu
Format: Others
Language:zh-TW
Published: 2016
Online Access:http://ndltd.ncl.edu.tw/handle/07063671262876028520
Description
Summary:碩士 === 中原大學 === 資訊工程研究所 === 104 === Shot Boundary Detection is a fundamental step of streaming video analysis and indexing. In the past decade, a lot of researches have been proposed for shot boundary detection. However, these methods are not effective if flashlight appeared in the videos. In this paper, a new mechanism is proposed to detect a frame with flashlight effect in streaming video. The distinctive character of flashlight effect in streaming video is the abrupt luminance change. When a flashlight occurs, the average luminance value goes to a higher value than the preceding frame and maintains one or several frames. Afterward, it returns back to original average luminance value. Such an abrupt luminance change will cause false detection of Shot Boundary. Based on this character, a mechanism to detect a frame with flashlight is designed. In this mechanism, differential value of luminance between this frame and adjacent frame is used to detect the video frame where flashlight appears. With this mechanism, the false detection rate of shot detection caused by flashlight can be reduced. The accuracy of our approach to detect flashlight scene is experimentally shown to be high. The Precision and Recall rate of the experiment result is around 0.9. The false detection of shot boundary detection is also effectively solved by this mechanism.