Robust Dissolve Detection Using Local Feature Tracking

In this paper we propose a novel dissolve detection algorithm which overcomes the defects of previously suggested algorithms. Differentiating dissolve from the effect aroused by camera motion or objects movement has long been a problem irritating the researchers and not settled thoroughly yet. In our approach, feature points are tracked so that inter-frame comparison can be executed between the exactly corresponding local features. This method reduces the false alarm caused by motions and improves the acuity of inter-frame comparison greatly. The experimental result shows the superiority of local feature comparison method to that of global feature