Fade and dissolve detection based on video editing model
基于视频编辑模型的视频淡入、淡出和叠化的检测

Based on video editing model, several important characteristics that can be used for detecting fade and dissolve, such as the variance, edge intensity, Effective Average Gradient (EAG) and Double Chromatic Difference (DCD) of video sequence are derived. After comparison, the variances are used to achieve robust fade detection. It is approved that in terms of recall rate for dissolve detection, the variance, edge intensity and EAG have similar efficiency. So, the variances are used for selection of possible positions of dissolve transitions, which guarantees high recall rate, and the DCD are then used for further confirmation to improve precision rate. Since the detection is carried out on DC sequence, the computation complexity of this method is relatively low.