Content Based Temporal Segmentation for Video Analysis
Narra Dhanalakshmi, Y. Madhavee Latha, Damodaram Avula · 2020
This temporal video segmentation aims at getting video segments so that the object's action should be enveloped by one segment. It is more challenging when each frame in the video is having multiple objects. Hence, it provides overlapped segments as a solution based on object-level information of the video rather than abrupt and gradual transitions. Temporal coherence between the content of frames is found to represent object-level information followed by a histogram intersection method to find a level of coincidence between the frames. The outcome of this work is suitable for many further video content analysis tasks such as understanding multimedia databases, tracking an object motion completely, efficient video content indexing, and so on. The segmentation method was tested on the sports genre which is one of the challenging domains because these video genres are composed of diverse scenes with multiple objects.