Scene Selection of Broadcast Soccer Video using Temporal Slices

In this paper, we propose a novel method for selecting personal favorite scenes using hierarchical eigenspace method based approach from broadcast soccer videos. In our method, we use physical parameters extracted from videos, such as image space frequency and time space frequency. We use these physical parameters as feature vectors, make an eigenspace of these vectors extracted from image slices of broadcasted soccer videos, and select the image slices that the distance from prior selected image slices as a personal favorite scene in the eigenspace is small. And we construct the hierarchical image pyramid which includes smaller resized images of input frames. Firstly we use the eigenspace of smallest resized frames of videos, and extract candidate frames of personal favorite scenes. In the next steps, we use the eigenspaces of next smallest resized frames. And finally we extract frames of personal favorite scenes using eigenspace of non-resized frames. The experimental results show the ability of our proposed method