Abstract
Due to the advancement of technology, a huge quantity of video data are captured and stored everyday all over the world for providing security, entertainment, sharing knowledge, and reporting news. It is a laborious, costly, and time consuming task to analyse, identify or investigate any significant events manually from a huge amount of video data. Furthermore, to store these gigantic video data, large memory space is also required. Therefore, it is essential to develop a method to extract important information from the huge volume of video data. Video Summarisation is a process to convert a long original video into a short duration containing all the important events (objects and their activities) without any repetition so that a user can comprehend the whole video quickly. However, it is still a challenging problem to extract objects and their activities from a video due to noise, illumination change, camera motion or shaking and low contrast. To address the challenges, we propose four innovative approaches for video summarisation in the highly diverse environments.
We introduce two new methods for video summarisation by extracting features from contents of a video. Firstly, a novel method is proposed using geometric primitives as these high level shape descriptors are extracted from edges which are prominent in the low contrast regions. Secondly, a new video summarisation technique is proposed by combining three multi-modal Human Visual Sensitive Features (HVSF) as the uni-modal feature cannot perform very well individually. We also propose two pioneering approaches to summarise a video by applying human perception during watching a video. We introduce a framework employing human Eye Tracker data as human eyes can track moving objects accurately in case of illumination change and camera movements. Furthermore, we apply Electroencephalography (EEG) signals to capture electrical activity in the human brain while watching a video as the human brain is the final evaluator of the video content comprehension. The experimental results reveal that the proposed four schemes outperform existing and relevant state-of-the-art methods.
We introduce two new methods for video summarisation by extracting features from contents of a video. Firstly, a novel method is proposed using geometric primitives as these high level shape descriptors are extracted from edges which are prominent in the low contrast regions. Secondly, a new video summarisation technique is proposed by combining three multi-modal Human Visual Sensitive Features (HVSF) as the uni-modal feature cannot perform very well individually. We also propose two pioneering approaches to summarise a video by applying human perception during watching a video. We introduce a framework employing human Eye Tracker data as human eyes can track moving objects accurately in case of illumination change and camera movements. Furthermore, we apply Electroencephalography (EEG) signals to capture electrical activity in the human brain while watching a video as the human brain is the final evaluator of the video content comprehension. The experimental results reveal that the proposed four schemes outperform existing and relevant state-of-the-art methods.
| Original language | English |
|---|---|
| Qualification | Doctor of Philosophy |
| Awarding Institution |
|
| Supervisors/Advisors |
|
| Award date | 28 Feb 2017 |
| Publication status | Published - 2017 |