Performance Evaluation of Video Segmentation Metrics

With the recent advancement of video data and daily creation of a large number of digital videos, the field of video indexing and retrieval is becoming an active area. For proper indexing and retrieval of the data, content-based video retrieval (CBVR) is a fundamental step for video segmentation. Content-based video retrieval is used for manipulating a video into frames. Shot detection is the fundamental step for automated indexing and content-based video retrieval. In this paper, the traditional metrics likelihood ratio, chi-square test, wavelet-based method are evaluated on datasets for detecting shot boundaries and measured in terms of recall, precision, and F1 measure. It has been determined that these results are affected by disturbance due to illumination and fire flicker.